Q1. (Given a Dataset) Analyze this dataset and give me a model that can predict this response variable.
Start by fitting a simple model (multivariate regression, logistic regression), do some feature engineering accordingly, and then try some complicated models. Always split the dataset into train, validation, test dataset and use cross validation to check their performance.
Determine if the problem is classification or regression
Favor simple models that run quickly and you can easily explain.
Mention cross validation as a means to evaluate the model.
Plot and visualize the data.
Q2. What could be some issues if the distribution of the test data is significantly different than the distribution of the training data?
The model that has high training accuracy might have low test accuracy. Without further knowledge, it is hard to know which dataset represents the population data and thus the generalizability of the algorithm is hard to measure. This should be mitigated by repeated splitting of train vs test dataset (as in cross validation).
When there is a change in data distribution, this is called the dataset shift. If the train and test data has a different distribution, then the classifier would likely overfit to the train data.
This issue can be overcome by using a more general learning method.
This can occur when:
P(y|x) are the same but P(x) are different. (covariate shift)
P(y|x) are different. (concept shift)
The causes can be:
Training samples are obtained in a biased way. (sample selection bias)
Train is different from test because of temporal, spatial changes. (non-stationary environments)
Solution to covariate shift
importance weighted cv
Q3. What are some ways I can make my model more robust to outliers?
We can have regularization such as L1 or L2 to reduce variance (increase bias).
Changes to the algorithm:
Use tree-based methods instead of regression methods as they are more resistant to outliers. For statistical tests, use non parametric tests instead of parametric ones.
Use robust error metrics such as MAE or Huber Loss instead of MSE.
Changes to the data:
Winsorizing the data
Transforming the data (e.g. log)
Remove them only if you’re certain they’re anomalies not worth predicting
Q4. What are some differences you would expect in a model that minimizes squared error, versus a model that minimizes absolute error? In which cases would each error metric be appropriate?
MSE is more strict to having outliers. MAE is more robust in that sense, but is harder to fit the model for because it cannot be numerically optimized. So when there are less variability in the model and the model is computationally easy to fit, we should use MAE, and if that’s not the case, we should use MSE.
MSE: easier to compute the gradient, MAE: linear programming needed to compute the gradient
MAE more robust to outliers. If the consequences of large errors are great, use MSE
MSE corresponds to maximizing likelihood of Gaussian random variables
Q5. What error metric would you use to evaluate how good a binary classifier is? What if the classes are imbalanced? What if there are more than 2 groups?
Accuracy: proportion of instances you predict correctly. Pros: intuitive, easy to explain, Cons: works poorly when the class labels are imbalanced and the signal from the data is weak
AUROC: plot fpr on the x axis and tpr on the y axis for different threshold. Given a random positive instance and a random negative instance, the AUC is the probability that you can identify who’s who. Pros: Works well when testing the ability of distinguishing the two classes, Cons: can’t interpret predictions as probabilities (because AUC is determined by rankings), so can’t explain the uncertainty of the model
logloss/deviance: Pros: error metric based on probabilities, Cons: very sensitive to false positives, negatives
When there are more than 2 groups, we can have k binary classifications and add them up for logloss. Some metrics like AUC is only applicable in the binary case.
Q6. What are various ways to predict a binary response variable? Can you compare two of them and tell me when one would be more appropriate? What’s the difference between these? (SVM, Logistic Regression, Naive Bayes, Decision Tree, etc.)
Things to look at: N, P, linearly seperable?, features independent?, likely to overfit?, speed, performance, memory usage
Logistic Regression
features roughly linear, problem roughly linearly separable
robust to noise, use l1,l2 regularization for model selection, avoid overfitting
the output come as probabilities
efficient and the computation can be distributed
can be used as a baseline for other algorithms
(-) can hardly handle categorical features
SVM
with a nonlinear kernel, can deal with problems that are not linearly separable
(-) slow to train, for most industry scale applications, not really efficient
Naive Bayes
computationally efficient when P is large by alleviating the curse of dimensionality
works surprisingly well for some cases even if the condition doesn’t hold
with word frequencies as features, the independence assumption can be seen reasonable. So the algorithm can be used in text categorization
(-) conditional independence of every other feature should be met
Tree Ensembles
good for large N and large P, can deal with categorical features very well
non parametric, so no need to worry about outliers
GBT’s work better but the parameters are harder to tune
RF works out of the box, but usually performs worse than GBT
Deep Learning
works well for some classification tasks (e.g. image)
used to squeeze something out of the problem.
Q7. What is regularization and where might it be helpful? What is an example of using regularization in a model?
Regularization is useful for reducing variance in the model, meaning avoiding overfitting . For example, we can use L1 regularization in Lasso regression to penalize large coefficients.
Q8. Why might it be preferable to include fewer predictors over many?
When we add irrelevant features, it increases model’s tendency to overfit because those features introduce more noise. When two variables are correlated, they might be harder to interpret in case of regression, etc.
curse of dimensionality
adding random noise makes the model more complicated but useless
computational cost
Ask someone for more details.
Q9. Given training data on tweets and their retweets, how would you predict the number of retweets of a given tweet after 7 days after only observing 2 days worth of data?
Build a time series model with the training data with a seven day cycle and then use that for a new data with only 2 days data.
Ask someone for more details.
Build a regression function to estimate the number of retweets as a function of time t
to determine if one regression function can be built, see if there are clusters in terms of the trends in the number of retweets
if not, we have to add features to the regression function
features + # of retweets on the first and the second day -> predict the seventh day
https://en.wikipedia.org/wiki/Dynamic_time_warping
Q10. How could you collect and analyze data to use social media to predict the weather?
We can collect social media data using twitter, Facebook, instagram API’s. Then, for example, for twitter, we can construct features from each tweet, e.g. the tweeted date, number of favorites, retweets, and of course, the features created from the tweeted content itself. Then use a multi variate time series model to predict the weather.
Q11. How would you construct a feed to show relevant content for a site that involves user interactions with items?
We can do so using building a recommendation engine. The easiest we can do is to show contents that are popular other users, which is still a valid strategy if for example the contents are news articles. To be more accurate, we can build a content based filtering or collaborative filtering. If there’s enough user usage data, we can try collaborative filtering and recommend contents other similar users have consumed. If there isn’t, we can recommend similar items based on vectorization of items (content based filtering).
Q12. How would you design the people you may know feature on LinkedIn or Facebook?
Find strong unconnected people in weighted connection graph
Define similarity as how strong the two people are connected
Given a certain feature, we can calculate the similarity based on
friend connections (neighbors)
Check-in’s people being at the same location all the time.
same college, workplace
Have randomly dropped graphs test the performance of the algorithm
ref. News Feed Optimization
Affinity score: how close the content creator and the users are
Weight: weight for the edge type (comment, like, tag, etc.). Emphasis on features the company wants to promote
Time decay: the older the less important
Q13. How would you predict who someone may want to send a Snapchat or Gmail to?
for each user, assign a score of how likely someone would send an email to
the rest is feature engineering:
number of past emails, how many responses, the last time they exchanged an email, whether the last email ends with a question mark, features about the other users, etc.
Ask someone for more details.
People who someone sent emails the most in the past, conditioning on time decay.
Q14. How would you suggest to a franchise where to open a new store?
build a master dataset with local demographic information available for each location.
local income levels, proximity to traffic, weather, population density, proximity to other businesses
a reference dataset on local, regional, and national macroeconomic conditions (e.g. unemployment, inflation, prime interest rate, etc.)
any data on the local franchise owner-operators, to the degree the manager
identify a set of KPIs acceptable to the management that had requested the analysis concerning the most desirable factors surrounding a franchise
quarterly operating profit, ROI, EVA, pay-down rate, etc.
run econometric models to understand the relative significance of each variable
run machine learning algorithms to predict the performance of each location candidate
Q15. In a search engine, given partial data on what the user has typed, how would you predict the user’s eventual search query?
Based on the past frequencies of words shown up given a sequence of words, we can construct conditional probabilities of the set of next sequences of words that can show up (n-gram). The sequences with highest conditional probabilities can show up as top candidates.
To further improve this algorithm,
we can put more weight on past sequences which showed up more recently and near your location to account for trends
show your recent searches given partial data
Q16. Given a database of all previous alumni donations to your university, how would you predict which recent alumni are most likely to donate?
Based on frequency and amount of donations, graduation year, major, etc, construct a supervised regression (or binary classification) algorithm.
Q17. You’re Uber and you want to design a heatmap to recommend to drivers where to wait for a passenger. How would you approach this?
Based on the past pickup location of passengers around the same time of the day, day of the week (month, year), construct
Ask someone for more details.
Based on the number of past pickups
account for periodicity (seasonal, monthly, weekly, daily, hourly)
special events (concerts, festivals, etc.) from tweets
Q18. How would you build a model to predict a March Madness bracket?
One vector each for team A and B. Take the difference of the two vectors and use that as an input to predict the probability that team A would win by training the model. Train the models using past tournament data and make a prediction for the new tournament by running the trained model for each round of the tournament
Some extensions:
Experiment with different ways of consolidating the 2 team vectors into one (e.g concantenating, averaging, etc)
Consider using a RNN type model that looks at time series data.
Q19. You want to run a regression to predict the probability of a flight delay, but there are flights with delays of up to 12 hours that are really messing up your model. How can you address this?
This is equivalent to making the model more robust to outliers.
See Q3.
Probability
Q1. Bobo the amoeba has a 25%, 25%, and 50% chance of producing 0, 1, or 2 o spring, respectively. Each of Bobo’s descendants also have the same probabilities. What is the probability that Bobo’s lineage dies out?
p=1/4+1/4p+1/2p^2 => p=1/2
Q2. In any 15-minute interval, there is a 20% probability that you will see at least one shooting star. What is the proba- bility that you see at least one shooting star in the period of an hour?
1-(0.8)^4. Or, we can use Poisson processes
Q3. How can you generate a random number between 1 – 7 with only a die?
Q4. How can you get a fair coin toss if someone hands you a coin that is weighted to come up heads more often than tails?
Flip twice and if HT then H, TH then T.
Q5. You have an 50-50 mixture of two normal distributions with the same standard deviation. How far apart do the means need to be in order for this distribution to be bimodal?
more than two standard deviations
Q6. Given draws from a normal distribution with known parameters, how can you simulate draws from a uniform distribution?
plug in the value to the CDF of the same random variable
Q7. A certain couple tells you that they have two children, at least one of which is a girl. What is the probability that they have two girls?
1/3
Q8. You have a group of couples that decide to have children until they have their first girl, after which they stop having children. What is the expected gender ratio of the children that are born?
What is the expected number of children each couple will have?
gender ratio is 1:1. Expected number of children is 2. let X be the number of children until getting a female (happens with prob 1/2). this follows a geometric distribution with probability 1/2
Q9. How many ways can you split 12 people into 3 teams of 4?
the outcome follows a multinomial distribution with n=12 and k=3. but the classes are indistinguishable
Q10. Your hash function assigns each object to a number between 1:10, each with equal probability. With 10 objects, what is the probability of a hash collision? What is the expected number of hash collisions? What is the expected number of hashes that are unused?
the probability of a hash collision: 1-(10!/10^10)
the expected number of hash collisions: 1-10*(9/10)^10
the expected number of hashes that are unused: 10*(9/10)^10
Q11. You call 2 UberX’s and 3 Lyfts. If the time that each takes to reach you is IID, what is the probability that all the Lyfts arrive first? What is the probability that all the UberX’s arrive first?
Lyfts arrive first: 2!*3!/5!
Ubers arrive first: same
Q12. I write a program should print out all the numbers from 1 to 300, but prints out Fizz instead if the number is divisible by 3, Buzz instead if the number is divisible by 5, and FizzBuzz if the number is divisible by 3 and 5. What is the total number of numbers that is either Fizzed, Buzzed, or FizzBuzzed?
100+60-20=140
Q13. On a dating site, users can select 5 out of 24 adjectives to describe themselves. A match is declared between two users if they match on at least 4 adjectives. If Alice and Bob randomly pick adjectives, what is the probability that they form a match?
24C5*(1+5(24-5))/24C5*24C5 = 4/1771
Q14. A lazy high school senior types up application and envelopes to n different colleges, but puts the applications randomly into the envelopes. What is the expected number of applications that went to the right college?
1
Q15. Let’s say you have a very tall father. On average, what would you expect the height of his son to be? Taller, equal, or shorter? What if you had a very short father?
Shorter. Regression to the mean
Q16. What’s the expected number of coin flips until you get two heads in a row? What’s the expected number of coin flips until you get two tails in a row?
Q17. Let’s say we play a game where I keep flipping a coin until I get heads. If the first time I get heads is on the nth coin, then I pay you 2n-1 dollars. How much would you pay me to play this game?
less than $3
Q18. You have two coins, one of which is fair and comes up heads with a probability 1/2, and the other which is biased and comes up heads with probability 3/4. You randomly pick coin and flip it twice, and get heads both times. What is the probability that you picked the fair coin?
4/13
Data Analysis
Q19. Let’s say you’re building the recommended music engine at Spotify to recommend people music based on past lis- tening history. How would you approach this problem?
collaborative filtering
Q20. What is R2? What are some other metrics that could be better than R2 and why?
goodness of fit measure. variance explained by the regression / total variance
the more predictors you add the higher R^2 becomes.
hence use adjusted R^2 which adjusts for the degrees of freedom
or train error metrics
Q21. What is the curse of dimensionality?
High dimensionality makes clustering hard, because having lots of dimensions means that everything is “far away” from each other.
For example, to cover a fraction of the volume of the data we need to capture a very wide range for each variable as the number of variables increases
All samples are close to the edge of the sample. And this is a bad news because prediction is much more difficult near the edges of the training sample.
The sampling density decreases exponentially as p increases and hence the data becomes much more sparse without significantly more data.
We should conduct PCA to reduce dimensionality
Q22. Is more data always better?
Statistically,
It depends on the quality of your data, for example, if your data is biased, just getting more data won’t help.
It depends on your model. If your model suffers from high bias, getting more data won’t improve your test results beyond a point. You’d need to add more features, etc.
Practically,
Also there’s a tradeoff between having more data and the additional storage, computational power, memory it requires. Hence, always think about the cost of having more data.
Q23. What are advantages of plotting your data before performing analysis?
Data sets have errors. You won’t find them all but you might find some. That 212 year old man. That 9 foot tall woman.
Variables can have skewness, outliers etc. Then the arithmetic mean might not be useful. Which means the standard deviation isn’t useful.
Variables can be multimodal! If a variable is multimodal then anything based on its mean or median is going to be suspect.
Q24. How can you make sure that you don’t analyze something that ends up meaningless?
Proper exploratory data analysis.
In every data analysis task, there’s the exploratory phase where you’re just graphing things, testing things on small sets of the data, summarizing simple statistics, and getting rough ideas of what hypotheses you might want to pursue further.
Then there’s the exploitatory phase, where you look deeply into a set of hypotheses.
The exploratory phase will generate lots of possible hypotheses, and the exploitatory phase will let you really understand a few of them. Balance the two and you’ll prevent yourself from wasting time on many things that end up meaningless, although not all.
Q25. What is the role of trial and error in data analysis? What is the role of making a hypothesis before diving in?
data analysis is a repetition of setting up a new hypothesis and trying to refute the null hypothesis.
The scientific method is eminently inductive: we elaborate a hypothesis, test it and refute it or not. As a result, we come up with new hypotheses which are in turn tested and so on. This is an iterative process, as science always is.
Q26. How can you determine which features are the most important in your model?
run the features though a Gradient Boosting Machine or Random Forest to generate plots of relative importance and information gain for each feature in the ensembles.
Look at the variables added in forward variable selection
Q27. How do you deal with some of your predictors being missing?
Remove rows with missing values – This works well if 1) the values are missing randomly (see Vinay Prabhu’s answer for more details on this) 2) if you don’t lose too much of the dataset after doing so.
Build another predictive model to predict the missing values – This could be a whole project in itself, so simple techniques are usually used here.
Use a model that can incorporate missing data – Like a random forest, or any tree-based method.
Q28. You have several variables that are positively correlated with your response, and you think combining all of the variables could give you a good prediction of your response. However, you see that in the multiple linear regression, one of the weights on the predictors is negative. What could be the issue?
Multicollinearity refers to a situation in which two or more explanatory variables in a multiple regression model are highly linearly related.
Leave the model as is, despite multicollinearity. The presence of multicollinearity doesn’t affect the efficiency of extrapolating the fitted model to new data provided that the predictor variables follow the same pattern of multicollinearity in the new data as in the data on which the regression model is based.
principal component regression
Q29. Let’s say you’re given an unfeasible amount of predictors in a predictive modeling task. What are some ways to make the prediction more feasible?
PCA
Q30. Now you have a feasible amount of predictors, but you’re fairly sure that you don’t need all of them. How would you perform feature selection on the dataset?
ridge / lasso / elastic net regression
Univariate Feature Selection where a statistical test is applied to each feature individually. You retain only the best features according to the test outcome scores
“Recursive Feature Elimination”:
First, train a model with all the feature and evaluate its performance on held out data.
Then drop let say the 10% weakest features (e.g. the feature with least absolute coefficients in a linear model) and retrain on the remaining features.
Iterate until you observe a sharp drop in the predictive accuracy of the model.
Q31. Your linear regression didn’t run and communicates that there are an infinite number of best estimates for the regression coefficients. What could be wrong?
p > n.
If some of the explanatory variables are perfectly correlated (positively or negatively) then the coefficients would not be unique.
Q32. You run your regression on different subsets of your data, and that in each subset, the beta value for a certain variable varies wildly. What could be the issue here?
The dataset might be heterogeneous. In which case, it is recommended to cluster datasets into different subsets wisely, and then draw different models for different subsets. Or, use models like non parametric models (trees) which can deal with heterogeneity quite nicely.
What is the main idea behind ensemble learning? If I had many different models that predicted the same response variable, what might I want to do to incorporate all of the models? Would you expect this to perform better than an individual model or worse?
The assumption is that a group of weak learners can be combined to form a strong learner.
Hence the combined model is expected to perform better than an individual model.
Assumptions:
average out biases
reduce variance
Bagging works because some underlying learning algorithms are unstable: slightly different inputs leads to very different outputs. If you can take advantage of this instability by running multiple instances, it can be shown that the reduced instability leads to lower error. If you want to understand why, the original bagging paper( http://www.springerlink.com/cont…) has a section called “why bagging works”
Boosting works because of the focus on better defining the “decision edge”. By reweighting examples near the margin (the positive and negative examples) you get a reduced error (see http://citeseerx.ist.psu.edu/vie…)
Use the outputs of your models as inputs to a meta-model.
For example, if you’re doing binary classification, you can use all the probability outputs of your individual models as inputs to a final logistic regression (or any model, really) that can combine the probability estimates.
One very important point is to make sure that the output of your models are out-of-sample predictions. This means that the predicted value for any row in your dataframe should NOT depend on the actual value for that row.
Q33. Given that you have wi data in your o ce, how would you determine which rooms and areas are underutilized and overutilized?
If the data is more used in one room, then that one is over utilized! Maybe account for the room capacity and normalize the data.
Q34. How would you quantify the influence of a Twitter user?
like page rank with each user corresponding to the web pages and linking to the page equivalent to following.
Q35. You have 100 mathletes and 100 math problems. Each mathlete gets to choose 10 problems to solve. Given data on who got what problem correct, how would you rank the problems in terms of difficulty?
One way you could do this is by storing a “skill level” for each user and a “difficulty level” for each problem. We assume that the probability that a user solves a problem only depends on the skill of the user and the difficulty of the problem.* Then we maximize the likelihood of the data to find the hidden skill and difficulty levels.
The Rasch model for dichotomous data takes the form:
{\displaystyle \Pr\{X_{ni}=1\}={\frac {\exp({\beta _{n}}-{\delta _{i}})}{1+\exp({\beta _{n}}-{\delta _{i}})}},}
where is the ability of person and is the difficulty of item}.
Q36. You have 5000 people that rank 10 sushis in terms of salt- iness. How would you aggregate this data to estimate the true saltiness rank in each sushi?
Some people would take the mean rank of each sushi. If I wanted something simple, I would use the median, since ranks are (strictly speaking) ordinal and not interval, so adding them is a bit risque (but people do it all the time and you probably won’t be far wrong).
Q37. Given data on congressional bills and which congressio- nal representatives co-sponsored the bills, how would you determine which other representatives are most similar to yours in voting behavior? How would you evaluate who is the most liberal? Most republican? Most bipartisan?
collaborative filtering. you have your votes and we can calculate the similarity for each representatives and select the most similar representative
for liberal and republican parties, find the mean vector and find the representative closest to the center point
Q38. How would you come up with an algorithm to detect plagiarism in online content?
reduce the text to a more compact form (e.g. fingerprinting,
bag of wor
ds) then compare those with other texts by calculating the similarity
Q39. You have data on all purchases of customers at a grocery store. Describe to me how you would program an algorithm that would cluster the customers into groups. How would you determine the appropriate number of clusters include?
KNN
choose a small value of k that still has a low SSE (elbow method)
https://bl.ocks.org/rpgove/0060ff3b656618e9136b
Statistical Inference
Q40. In an A/B test, how can you check if assignment to the various buckets was truly random?
Plot the distributions of multiple features for both A and B and make sure that they have the same shape. More rigorously, we can conduct a permutation test to see if the distributions are the same.
MANOVA to compare different means
Q41. What might be the benefits of running an A/A test, where you have two buckets who are exposed to the exact same product?
Verify the sampling algorithm is random.
Q42. What would be the hazards of letting users sneak a peek at the other bucket in an A/B test?
The user might not act the same suppose had they not seen the other bucket. You are essentially adding additional variables of whether the user peeked the other bucket, which are not random across groups.
Q43. What would be some issues if blogs decide to cover one of your experimental groups?
Same as the previous question. The above problem can happen in larger scale.
Q44. How would you conduct an A/B test on an opt-in feature?
Ask someone for more details.
Q45. How would you run an A/B test for many variants, say 20 or more?
one control, 20 treatment, if the sample size for each group is big enough.
Ways to attempt to correct for this include changing your confidence level (e.g. Bonferroni Correction) or doing family-wide tests before you dive in to the individual metrics (e.g. Fisher’s Protected LSD).
Q46. How would you run an A/B test if the observations are extremely right-skewed?
lower the variability by modifying the KPI
cap values
percentile metrics
log transform
https://www.quora.com/How-would-you-run-an-A-B-test-if-the-observations-are-extremely-right-skewed
Q47. I have two different experiments that both change the sign-up button to my website. I want to test them at the same time. What kinds of things should I keep in mind?
exclusive -> ok
Q48. What is a p-value? What is the difference between type-1 and type-2 error?
type-1 error: rejecting Ho when Ho is a true
type-2 error: not rejecting Ho when Ha is true
Q49. You are AirBnB and you want to test the hypothesis that a greater number of photographs increases the chances that a buyer selects the listing. How would you test this hypothesis?
For randomly selected listings with more than 1 pictures, hide 1 random picture for group A, and show all for group B. Compare the booking rate for the two groups.
Ask someone for more details.
Q50. How would you design an experiment to determine the impact of latency on user engagement?
The best way I know to quantify the impact of performance is to isolate just that factor using a slowdown experiment, i.e., add a delay in an A/B test.
Q51. What is maximum likelihood estimation? Could there be any case where it doesn’t exist?
A method for parameter optimization (fitting a model). We choose parameters so as to maximize the likelihood function (how likely the outcome would happen given the current data and our model).
maximum likelihood estimation (MLE) is a method of estimating the parameters of a statistical model given observations, by finding the parameter values that maximize the likelihood of making the observations given the parameters. MLE can be seen as a special case of the maximum a posteriori estimation (MAP) that assumes a uniform prior distribution of the parameters, or as a variant of the MAP that ignores the prior and which therefore is unregularized.
for Gaussian mixtures, non-parametric models, it doesn’t exist
Q52. What’s the difference between a MAP, MOM, MLE estimator? In which cases would you want to use each?
MAP estimates the posterior distribution given the prior distribution and data which maximizes the likelihood function. MLE is a special case of MAP where the prior is uninformative uniform distribution.
MOM sets moment values and solves for the parameters. MOM has not used much anymore because maximum likelihood estimators have higher probability of being close to the quantities to be estimated and are more often unbiased.
Q53. What is a confidence interval and how do you interpret it?
For example, 95% confidence interval is an interval that when constructed for a set of samples each sampled in the same way, the constructed intervals include the true mean 95% of the time.
if confidence intervals are constructed using a given confidence level in an infinite number of independent experiments, the proportion of those intervals that contain the true value of the parameter will match the confidence level.
Q54. What is unbiasedness as a property of an estimator? Is this always a desirable property when performing inference? What about in data analysis or predictive modeling?
Unbiasedness means that the expectation of the estimator is equal to the population value we are estimating. This is desirable in inference because the goal is to explain the dataset as accurately as possible. However, this is not always desirable for data analysis or predictive modeling as there is the bias variance tradeoff. We sometimes want to prioritize the generalizability and avoid overfitting by reducing variance and thus increasing bias.
OTHER Important Data Science Interview Questions and Answers
Q55. What is the difference between population and sample in data?
Sample is the set of people who participated in your study whereas the population is the set of people to whom you want to generalize the results. For example – If you want to study the obesity among the children in India and you study 1000 children then those 1000 became sample whereas the all the children in the country is the population.
Sample is the subset of population.
Q56. What is the difference sample and sample frame?
Sample frame is the number of people who wanted to study whereas sample is the actual number of people who participated in your study. Ex – If you sent a marketing survey link to 300 people through email and only 100 participated in the survey then 300 is the sample survey and 100 is the sample.
Sample is the subset of sample frame. Both Sample and Sample Frame are subset of population.
Q57. What is the difference between univariate, bivariate and multivariate analysis?
Univariate analysis is performed on one variable, bivariate on two variable and multivariate analysis on two or more variables
Q58. What is difference between interpolation and extrapolation?
Extrapolation is the estimation of future values based on the observed trend on the past. Interpolation is the estimation of missing past values within two values in a sequence of values
Q59. What is precision and recall?
Precision is the percentage of correct predictions you have made and recall is the percentage of predictions that actually turned out to be true
Q60. What is confusion matrix?
Confusion matrix is a table which contains information about predicted values and actual values in a classification model
It has four parts namely true positive ,true negative, false positive and false negative
It can be used to calculate accuracy, precision and recall
Q61. What is hypothesis testing?
While performing the an experiment hypothesis testing to is used to analyze the various factors that are assumed to have an impact on the outcome of experiment
An hypothesis is some kind of assumption and hypothesis testing is used to determine whether the stated hypothesis is true or not
Initial assumption is called null hypothesis and the opposite alternate hypothesis
Q62. What is a p-value in statistics?
In hypothesis testing, p value helps to arrive at a conclusion. When p -value is too small then null hypothesis is rejected and alternate is accepted. When p-value is large then null hypothesis is accepted.
Q63. What is difference between Type-I error and Type-II error in hypothesis testing?
Type-I error is we reject the null hypothesis which was supposed to be accepted. It represents false positive
Type-II error represents we accept the null hypothesis which was supposed to be rejected. It represents false negative.
Q64. QWhat are the different types of missing value treatment?
Deletion of values
Guess the value
Average Substitution
Regression based substitution
Multiple Imputation
Q65. What is gradient descent?
When building a statistical model the objective is reduce the value of the cost function that is associated with the model. Gradient descent is an iterative optimization technique used to determine the minima of the cost function
Q66. What is difference between supervised and unsupervised learning algorithms?
Supervised learning are the class of algorithms in which model is trained by explicitly labelling the outcome. Ex. Regression, Classification
Unsupervised learning no output is given and the algorithm is made to learn the outcomes implicity Ex. Association, Clustering
Q67. What is the need for regularization in model building?
Regularization is used to penalize the model when it overfits the model. It predominantly helps in solving the overfitting problem.
Q68. Difference between bias and variance tradeoff?
High Bias is an underlying error wrong assumption that makes the model to underfit. High Variance in a model means noise in data has been too taken seriously by the model which will result in overfitting.
Typically we would like to have a model with low bias and low variance
Q69. How to solve overfitting?
Introduce Regularization
Perform Cross Validation
Reduce the number of features
Increase the number of entries
Ensembling
Q70. How will you detect the presence of overfitting?
When you build a model which has very high model accuracy on train data set and very low prediction accuracy in test data set then it is a indicator of overfitting
Q71. How do you determine the number of clusters in k-means clustering?
Elbow method ( Plotting the percentage of variance explained w.r.t to number of clusters)
Gap Statistic
Silhouette method
Q72. What is the difference between causality and correlation?
Correlation is the measure that helps us understand the relationship between two or more variables
Causation represents that causal relationship between two events. It is also known to represent cause and effect
Causation means there is correlation but correlation doesn’t necessarily mean causation
Q73. Explain normal distribution?
Normal distribution is a bell shaped curve that represents distribution of data around its mean. Any normal process would follow the normal distribution.
Most of data points tend to concentrated around the mean. If a point is further away from the mean then it is less likely to appear
Q74. What are the different ways of performing aggregation in python using pandas?
Group by function
Pivot function
Aggregate function
Q75. What are merge two list and get only unique values?
List a = [1,2,3,4] List b= [1,2,5,6] A = list(set(a+b))
Q76. How to save and retrieve model objects in python?
By using a library called pickle you can train any model and store the object in a pickle file.
When needed in future you can retrieve the object and use the model for prediction.
Q77. What is an anomaly and how is it different from outliers?
Anomaly detection is identification of items or events that didn’t fit to the exact pattern or other items in a dataset. Outliers are valid data points that are outside the norm whereas anomaly are invalid data points that are created by process that is different from process that created the other data points
Q78. What is an ensemble learning?
Ensemble learning is the art of combining more than one model to predict the final outcome of an experiment. Commonly used ensemble techniques bagging, boosting and stacking
Q79. Name few libraries that is used in python for data analysis?
Numpy
Scipy
Pandas
Scikit learn
Matplotlib\ seaborn
Q80. What are the different types of data?
Data is broadly classified into two types 1) Numerical 2) Categorical
Numerical variables is further classified into discrete and continuous data
Categorical variables
Systematic Sampling
Stratified Sampling
Quota Sampling are further classified into Binary, Nominal and Ordinal data
Q81. What is a lambda function in python?
Lambda function are used to create small, one-time anonymous function in python. It enables the programmer to create functions without a name and almost instantly
Q82. What are the different sampling methods?
Random Sampling
Systematic Sampling
Stratified Sampling
Quota Sampling
Q83. Common Data Quality Issues
Missing Values
Noise in the Data Set
Outliers
Mixture of Different Languages (like English and Chinese)
Range Constraints
Q84. What is the difference between supervised learning and un-supervised learning?
Supervised learning: Target variable is available and the algorithm learns for the train data
And applies to test data (unseen data).
Unsupervised learning: Target variable is not available and the algorithm does not need to learn
Anything beforehand.
Q85. What is Imbalanced Data Set and how to handle them? Name Few Examples?
Fraud detection
Disease screening
Imbalanced Data Set means that the population of one class is extremely large than the other
(Eg: Fraud – 99% and Non-Fraud – 1%)
Imbalanced dataset can be handled by either oversampling, undersampling and penalized Machine Learning Algorithm.
Q86. If you are dealing with 10M Data, then will you go for Machine learning (or) Deep learning Algorithm?
Machine learning algorithm suits well for small data and it might take huge amount of time to train for large data.
Whereas Deep learning algorithm takes less amount of data to train due to the help of GPU(Parallel Processing).
Q87. Examples of Supervised learning algorithm?
Linear Regression and Logistic Regression
Decision Trees and Random Forest
SVM
Naïve Bayes
XGBoost
Q88. In Logistic Regression, if you want to know the best features in your dataset then what you would do?
Apply step function, which calculates the AIC for different permutation and combination of features and provides the best features for the dataset.
Q89. What is Feature Engineering? Explain with Example?
Feature engineering is the process of using domain knowledge of the data to create features for machine learning algorithm to work
Adding more columns (or) removing columns from the existing column
Outlier Detection
Normalization etc
Q90. How to select the important features in the given data set?
In Logistic Regression, we can use step() which gives AIC score of set of features
In Decision Tree, We can use information gain(which internally uses entropy)
In Random Forest, We can use varImpPlot
Q91. When does multicollinearity problem occur and how to handle it?
It exists when 2 or more predictors are highly correlated with each other.
Example: In the Data Set if you have grades of 2nd PUC and marks of 2nd PUC, Then both gives the same trend to capture, which might internally hamper the speed and time.so we need to check if the multi collinearity exists by using VIF(variance Inflation Factor).
Note: if the Variance Inflation Factor is more than 4, then multi collinearity problem exists.
Q92. What is Variance inflation Factors (VIF)
Measure how much the variance of the estimated regression coefficients are inflated as compared to when the predictor variables are not linearly related.
Q93. Examples of Parametric machine learning algorithm and non-parametric machine learning algorithm
Parametric machine learning algorithm – Linear Regression, Logistic Regression
Non-Parametric machine learning algorithm – Decision Trees, SVM, Neural Network
Q94. What are parametric and non-parametric machine learning algorithm? And their importance
Algorithm which does not make strong assumptions are non-parametric algorithm and they are free to learn from training data.
Algorithm that makes strong assumptions are parametric and it involves 1) select the form for the function and 2)learn the coefficients for the function from training data.
Q95. When does linear and logistic regression performs better, generally?
It works better when we remove the attributes which are unrelated to the output variable and highly co-related variable to each other.
Q96. Why you call naïve bayes as “naïve” ?
Reason: It assumes that the input variable is independent, but in real world it is unrealistic, since all the features would be dependent on each other.
Q97. Give some example for false positive, false negative, true positive, true negative
False Positive – A cancer screening test comes back positive, but you don’t have cancer
False Negative – A cancer screening test comes back negative, but you have cancer
True Positive – A Cancer Screening test comes back positive, and you have cancer
True Negative – A Cancer Screening test comes back negative, and you don’t have cancer
Q98. What is Sensitivity and Specificity?
Sensitivity means “proportion of actual positives that are correctly classified” in other words “True Positive”
Specificity means “proportion of actual negatives that are correctly classified” “True Negative”
Q99. When to use Logistic Regression and when to use Linear Regression?
If you are dealing with a classification problem like (Yes/No, Fraud/Non Fraud, Sports/Music/Dance) then use Logistic Regression.
If you are dealing with continuous/discrete values, then go for Linear Regression.
Q100. What are the different imputation algorithm available?
Imputation algorithm means “replacing the Blank values by some values)
Mean imputation
Median Imputation
MICE
miss forest
Amelia
Q101. What is AIC(Akaike Information Criteria)
The analogous metric of adjusted R² in logistic regression is AIC.
AIC is the measure of fit which penalizes model for the number of model coefficients. Therefore, we always prefer model with minimum AIC value.
Q102. Suppose you have 10 samples, where 8 are positive and 2 are negative, how to calculate Entropy (important to know)
E(S) = 8/10log(8/10) – 2/10log(2/10)
Note: Log is à base 2
Q103. What is perceptron in Machine Leaning?
In Machine Learning. Perceptron is an algorithm for supervised classification of the input into one of several possible non-binary outputs
Q104. How to ensure we are not over fitting the model?
Keep the attributes/Columns which are really important
Use K-Fold cross validation techniques
Make use of drop-put incase of neural network
Q105. How the root node is predicted in Decision Tree Algorithm?
Mathematical Formula “Entropy” is utilized for predicting the root node of the tree.
Q106. What are the different Backend Process available in Keras?
TensorFlow
Theano
CNTK
Q107. Name Few Deep Learning Algorithm
TensorFlow
Theano
Lasagne
mxnet
blocks
Keras
CNTK
TFLearn
Q108. How to split the data with equal set of classes in both training and testing data?
Using Stratified Shuffle package
Q109. What do you mean by giving “epoch = 1” in neural network?
It means that “traversing the data set one time
Q110. What do you mean by Ensemble Model? When to use?
Ensemble Model is a combination of Different Models to predict correctly and with good accuracy.
Ensemble learning is used when you build component classifiers that are more accurate and independent from each other.
Q111. When will you use SVM and when to use Random Forest?
SVM can be used if the data is outlier free whereas Naïve Bayes can be used even if it has outliers (since it has built in package to take care).
SVM suits best for Text Classification Model and Random Forest suits for Binomial/Multinomial Classification Problem.
Random Forest takes care of over fitting problem with the help of tree pruning
Q112. Applications of Machine Learning?
Self Driving Cars
Image Classification
Text Classification
Search Engine
Banking, Healthcare Domain
Q113. If you are given with a use case – ‘Predict whether the transaction is fraud (or) not fraud”, which algorithm would you choose
Logistic Regression
Q114. If you are given with a use case – ‘Predict the house price range in the coming years”, which algorithm would you choose
Linear Regression
Q115. What is the underlying mathematical knowledge behind Naïve Bayes?
Bayes Theorem
Q116. When to use Random Forest and when to Use XGBoost?
If you want all core processors in your system to be utilized, then go for XGBoost(since it supports parallel processing) and if your data is small then go for random forest.
Q117. If you are training model gives 90% accuracy and test model gives 60% accuracy? Then what problem you are facing with?
Overfitting.
Overfitting and can be reduced by many methods like (Tree Pruning, Removing the minute information provided in the data set).
Q118. In Google if you type “How are “it gives you the recommendation as “How are you “/”How do you do”, this is based on what?
This kind of recommendation engine comes from collaborative filtering.
Q120. What is Boosting? Explain how Boosting works?
Boosting is a Ensemble technique that attempts to create strong classifier from a number of weak classifiers
After the first tree is created, the performance of the tree on each training instance is used to weight how much attention the next tree that is created should pay attention to each training instance by giving more weights to the misclassified one.
Models are created one after the other, each updating the weights on the training instance
Q121. What is Null Deviance and Residual Deviance (Logistic Regression Concept?)
Null Deviance indicates the response predicted by a model with nothing but an intercept
Residual deviance indicates the response predicted by a model on adding independent variables
Note:
Lower the value, better the model
Q122. What are the different method to split the tree in decision tree?
Information gain and gini index
Q123. What is the weakness for Decision Tree Algorithm?
Not suitable for continuous/Discrete variable
Performs poorly on small data
Q124. Why do we use PCA(Principal Components Analysis) ?
These are important feature extraction techniques used for dimensionality reduction.
Q125. During Imbalanced Data Set, will you
Calculate the Accuracy only? (or)
Precision, Recall, F1 Score separately
We need to calculate precision, Recall separately
Q126. How to ensure we are not over fitting the model?
Keep the attributes/Columns which are really important
Use K-Fold cross validation techniques
make use of drop-put in case of neural network
Q127. Steps involved in Decision Tree and finding the root node for the tree
Step 1:- How to find the Root Node
Use Information gain to understand the each attribute information w.r.t target variable and place the attribute with the highest information gain as root node.
Step 2:- How to Find the Information Gain
Please apply the entropy (Mathematical Formulae) to calculate Information Gain. Gain (T,X) = Entropy(T) – Entropy(T,X) here represent target variable and X represent features.
Step3: Identification of Terminal Node
Based on the information gain value obtained from the above steps, identify the second most highest information gain and place it as the terminal node.
Step 4: Predicted Outcome
Recursively iterate the step4 till we obtain the leaf node which would be our predicted target variable.
Step 5: Tree Pruning and optimization for good results
It helps to reduce the size of decision trees by removing sections of the tree to avoid over fitting.
Q128. What is hyper plane in SVM?
It is a line that splits the input variable space and it is selected to best separate the points in the input variable space by their class(0/1,yes/no).
Q129. Explain Bigram with an Example?
Eg: I Love Data Science
Bigram – (I Love) (Love Data) (Data Science)
Q130. What are the different activation functions in neural network?
Relu, Leaky Relu , Softmax, Sigmoid
Q131. Which Algorithm Suits for Text Classification Problem?
SVM, Naïve Bayes, Keras, Theano, CNTK, TFLearn(Tensorflow)
Q132. You are given a train data set having lot of columns and rows. How do you reduce the dimension of this data?
Principal Component Analysis(PCA) would help us here which can explain the maximum variance in the data set.
We can also check the co-relation for numerical data and remove the problem of multi-collinearity(if exists) and remove some of the columns which may not impact the model.
We can create multiple dataset and execute them batch wise.
Q133. You are given a data set on fraud detection. Classification model achieved accuracy of 95%.Is it good?
Accuracy of 96% is good. But we may have to check the following items:
what was the dataset for the classification problem
Is Sensitivity and Specificity are acceptable
if there are only less negative cases, and all negative cases are not correctly classified, then it might be a problem
In-Addition it is related to fraud detection, hence needs to be careful here in prediction (i.e not wrongly predicting the fraud as non-fraud patient.
Q134. What is prior probability and likelihood?
Prior probability:
The proportion of dependent variable in the data set.
Likelihood:
It is the probability of classifying a given observation as ‘1’ in the presence of some other variable.
Q135. How can we know if your data is suffering from low bias and high variance?
Random Forest Algorithm can be used to tackle high variance problem.in the cases of low bias and high variance L1,L2 regularization can help.
Q134. How is kNN different from kmeans clustering?
Kmeans partitions a data set into clusters, which is homogeneous and points in the cluster are close to each other. Whereas KNN tries to classify unlabelled observation based on its K surrounding neighbours.
Q135. Random Forest has 1000 trees, Training error: 0.0 and validation error is 20.00.What is the issue here?
It is the classical example of over fitting. It is not performing well on the unseen data. We may have to tune our model using cross validation and other techniques to overcome over fitting
Q136. Data set consisting of variables having more than 30% missing values? How will you deal with them?
We can remove them, if it does not impact our model
We can apply imputation techniques (like MICE, MISSFOREST,AMELIA) to avoid missing values
Q137. What do you understand by Type I vs. Type II error?
Type I error occurs when – “we classify a value as positive, when the actual value is negative”
(False Positive)
Type II error occurs when – “we classify a value as negative, when the actual value if positive”
(False Negative)
Q138. Based on the dataset, how will you know which algorithm to apply ?
If it is classification related problem,then we can use logistic,decision trees etc…
If it is Regression related problem, then we can use Linear Regression.
If it is Clustering based, we can use KNN.
We can also apply XGB, RF for better accuracy.
Q139. Why normalization is important?
Data Set can have one column in the range (10,000/20,000) and other column might have data in the range (1, 2, 3).clearly these two columns are in different range and cannot accurately analyse the trend. So we can apply normalization here by using min-max normalization (i.e to convert it into 0-1 scale).
Start by fitting a simple model (multivariate regression, logistic regression), do some feature engineering accordingly, and then try some complicated models. Always split the dataset into train, validation, test dataset and use cross validation to check their performance.
Determine if the problem is classification or regression
Favor simple models that run quickly and you can easily explain.
Mention cross validation as a means to evaluate the model.
Plot and visualize the data.
Q2. What could be some issues if the distribution of the test data is significantly different than the distribution of the training data?
The model that has high training accuracy might have low test accuracy. Without further knowledge, it is hard to know which dataset represents the population data and thus the generalizability of the algorithm is hard to measure. This should be mitigated by repeated splitting of train vs test dataset (as in cross validation).
When there is a change in data distribution, this is called the dataset shift. If the train and test data has a different distribution, then the classifier would likely overfit to the train data.
This issue can be overcome by using a more general learning method.
This can occur when:
P(y|x) are the same but P(x) are different. (covariate shift)
P(y|x) are different. (concept shift)
The causes can be:
Training samples are obtained in a biased way. (sample selection bias)
Train is different from test because of temporal, spatial changes. (non-stationary environments)
Solution to covariate shift
importance weighted cv
Q3. What are some ways I can make my model more robust to outliers?
We can have regularization such as L1 or L2 to reduce variance (increase bias).
Changes to the algorithm:
Use tree-based methods instead of regression methods as they are more resistant to outliers. For statistical tests, use non parametric tests instead of parametric ones.
Use robust error metrics such as MAE or Huber Loss instead of MSE.
Changes to the data:
Winsorizing the data
Transforming the data (e.g. log)
Remove them only if you’re certain they’re anomalies not worth predicting
Q4. What are some differences you would expect in a model that minimizes squared error, versus a model that minimizes absolute error? In which cases would each error metric be appropriate?
MSE is more strict to having outliers. MAE is more robust in that sense, but is harder to fit the model for because it cannot be numerically optimized. So when there are less variability in the model and the model is computationally easy to fit, we should use MAE, and if that’s not the case, we should use MSE.
MSE: easier to compute the gradient, MAE: linear programming needed to compute the gradient
MAE more robust to outliers. If the consequences of large errors are great, use MSE
MSE corresponds to maximizing likelihood of Gaussian random variables
Q5. What error metric would you use to evaluate how good a binary classifier is? What if the classes are imbalanced? What if there are more than 2 groups?
Accuracy: proportion of instances you predict correctly. Pros: intuitive, easy to explain, Cons: works poorly when the class labels are imbalanced and the signal from the data is weak
AUROC: plot fpr on the x axis and tpr on the y axis for different threshold. Given a random positive instance and a random negative instance, the AUC is the probability that you can identify who’s who. Pros: Works well when testing the ability of distinguishing the two classes, Cons: can’t interpret predictions as probabilities (because AUC is determined by rankings), so can’t explain the uncertainty of the model
logloss/deviance: Pros: error metric based on probabilities, Cons: very sensitive to false positives, negatives
When there are more than 2 groups, we can have k binary classifications and add them up for logloss. Some metrics like AUC is only applicable in the binary case.
Q6. What are various ways to predict a binary response variable? Can you compare two of them and tell me when one would be more appropriate? What’s the difference between these? (SVM, Logistic Regression, Naive Bayes, Decision Tree, etc.)
Things to look at: N, P, linearly seperable?, features independent?, likely to overfit?, speed, performance, memory usage
Logistic Regression
features roughly linear, problem roughly linearly separable
robust to noise, use l1,l2 regularization for model selection, avoid overfitting
the output come as probabilities
efficient and the computation can be distributed
can be used as a baseline for other algorithms
(-) can hardly handle categorical features
SVM
with a nonlinear kernel, can deal with problems that are not linearly separable
(-) slow to train, for most industry scale applications, not really efficient
Naive Bayes
computationally efficient when P is large by alleviating the curse of dimensionality
works surprisingly well for some cases even if the condition doesn’t hold
with word frequencies as features, the independence assumption can be seen reasonable. So the algorithm can be used in text categorization
(-) conditional independence of every other feature should be met
Tree Ensembles
good for large N and large P, can deal with categorical features very well
non parametric, so no need to worry about outliers
GBT’s work better but the parameters are harder to tune
RF works out of the box, but usually performs worse than GBT
Deep Learning
works well for some classification tasks (e.g. image)
used to squeeze something out of the problem.
 |
| Data Science Experienced Level Interview Questions Answers |
Q7. What is regularization and where might it be helpful? What is an example of using regularization in a model?
Regularization is useful for reducing variance in the model, meaning avoiding overfitting . For example, we can use L1 regularization in Lasso regression to penalize large coefficients.
Q8. Why might it be preferable to include fewer predictors over many?
When we add irrelevant features, it increases model’s tendency to overfit because those features introduce more noise. When two variables are correlated, they might be harder to interpret in case of regression, etc.
curse of dimensionality
adding random noise makes the model more complicated but useless
computational cost
Ask someone for more details.
Q9. Given training data on tweets and their retweets, how would you predict the number of retweets of a given tweet after 7 days after only observing 2 days worth of data?
Build a time series model with the training data with a seven day cycle and then use that for a new data with only 2 days data.
Ask someone for more details.
Build a regression function to estimate the number of retweets as a function of time t
to determine if one regression function can be built, see if there are clusters in terms of the trends in the number of retweets
if not, we have to add features to the regression function
features + # of retweets on the first and the second day -> predict the seventh day
https://en.wikipedia.org/wiki/Dynamic_time_warping
Q10. How could you collect and analyze data to use social media to predict the weather?
We can collect social media data using twitter, Facebook, instagram API’s. Then, for example, for twitter, we can construct features from each tweet, e.g. the tweeted date, number of favorites, retweets, and of course, the features created from the tweeted content itself. Then use a multi variate time series model to predict the weather.
Q11. How would you construct a feed to show relevant content for a site that involves user interactions with items?
We can do so using building a recommendation engine. The easiest we can do is to show contents that are popular other users, which is still a valid strategy if for example the contents are news articles. To be more accurate, we can build a content based filtering or collaborative filtering. If there’s enough user usage data, we can try collaborative filtering and recommend contents other similar users have consumed. If there isn’t, we can recommend similar items based on vectorization of items (content based filtering).
Q12. How would you design the people you may know feature on LinkedIn or Facebook?
Find strong unconnected people in weighted connection graph
Define similarity as how strong the two people are connected
Given a certain feature, we can calculate the similarity based on
friend connections (neighbors)
Check-in’s people being at the same location all the time.
same college, workplace
Have randomly dropped graphs test the performance of the algorithm
ref. News Feed Optimization
Affinity score: how close the content creator and the users are
Weight: weight for the edge type (comment, like, tag, etc.). Emphasis on features the company wants to promote
Time decay: the older the less important
Q13. How would you predict who someone may want to send a Snapchat or Gmail to?
for each user, assign a score of how likely someone would send an email to
the rest is feature engineering:
number of past emails, how many responses, the last time they exchanged an email, whether the last email ends with a question mark, features about the other users, etc.
Ask someone for more details.
People who someone sent emails the most in the past, conditioning on time decay.
Q14. How would you suggest to a franchise where to open a new store?
build a master dataset with local demographic information available for each location.
local income levels, proximity to traffic, weather, population density, proximity to other businesses
a reference dataset on local, regional, and national macroeconomic conditions (e.g. unemployment, inflation, prime interest rate, etc.)
any data on the local franchise owner-operators, to the degree the manager
identify a set of KPIs acceptable to the management that had requested the analysis concerning the most desirable factors surrounding a franchise
quarterly operating profit, ROI, EVA, pay-down rate, etc.
run econometric models to understand the relative significance of each variable
run machine learning algorithms to predict the performance of each location candidate
Q15. In a search engine, given partial data on what the user has typed, how would you predict the user’s eventual search query?
Based on the past frequencies of words shown up given a sequence of words, we can construct conditional probabilities of the set of next sequences of words that can show up (n-gram). The sequences with highest conditional probabilities can show up as top candidates.
To further improve this algorithm,
we can put more weight on past sequences which showed up more recently and near your location to account for trends
show your recent searches given partial data
Q16. Given a database of all previous alumni donations to your university, how would you predict which recent alumni are most likely to donate?
Based on frequency and amount of donations, graduation year, major, etc, construct a supervised regression (or binary classification) algorithm.
Q17. You’re Uber and you want to design a heatmap to recommend to drivers where to wait for a passenger. How would you approach this?
Based on the past pickup location of passengers around the same time of the day, day of the week (month, year), construct
Ask someone for more details.
Based on the number of past pickups
account for periodicity (seasonal, monthly, weekly, daily, hourly)
special events (concerts, festivals, etc.) from tweets
Q18. How would you build a model to predict a March Madness bracket?
One vector each for team A and B. Take the difference of the two vectors and use that as an input to predict the probability that team A would win by training the model. Train the models using past tournament data and make a prediction for the new tournament by running the trained model for each round of the tournament
Some extensions:
Experiment with different ways of consolidating the 2 team vectors into one (e.g concantenating, averaging, etc)
Consider using a RNN type model that looks at time series data.
Q19. You want to run a regression to predict the probability of a flight delay, but there are flights with delays of up to 12 hours that are really messing up your model. How can you address this?
This is equivalent to making the model more robust to outliers.
See Q3.
Probability
Q1. Bobo the amoeba has a 25%, 25%, and 50% chance of producing 0, 1, or 2 o spring, respectively. Each of Bobo’s descendants also have the same probabilities. What is the probability that Bobo’s lineage dies out?
p=1/4+1/4p+1/2p^2 => p=1/2
Q2. In any 15-minute interval, there is a 20% probability that you will see at least one shooting star. What is the proba- bility that you see at least one shooting star in the period of an hour?
1-(0.8)^4. Or, we can use Poisson processes
Q3. How can you generate a random number between 1 – 7 with only a die?
Q4. How can you get a fair coin toss if someone hands you a coin that is weighted to come up heads more often than tails?
Flip twice and if HT then H, TH then T.
Q5. You have an 50-50 mixture of two normal distributions with the same standard deviation. How far apart do the means need to be in order for this distribution to be bimodal?
more than two standard deviations
Q6. Given draws from a normal distribution with known parameters, how can you simulate draws from a uniform distribution?
plug in the value to the CDF of the same random variable
Q7. A certain couple tells you that they have two children, at least one of which is a girl. What is the probability that they have two girls?
1/3
Q8. You have a group of couples that decide to have children until they have their first girl, after which they stop having children. What is the expected gender ratio of the children that are born?
What is the expected number of children each couple will have?
gender ratio is 1:1. Expected number of children is 2. let X be the number of children until getting a female (happens with prob 1/2). this follows a geometric distribution with probability 1/2
Q9. How many ways can you split 12 people into 3 teams of 4?
the outcome follows a multinomial distribution with n=12 and k=3. but the classes are indistinguishable
Q10. Your hash function assigns each object to a number between 1:10, each with equal probability. With 10 objects, what is the probability of a hash collision? What is the expected number of hash collisions? What is the expected number of hashes that are unused?
the probability of a hash collision: 1-(10!/10^10)
the expected number of hash collisions: 1-10*(9/10)^10
the expected number of hashes that are unused: 10*(9/10)^10
Q11. You call 2 UberX’s and 3 Lyfts. If the time that each takes to reach you is IID, what is the probability that all the Lyfts arrive first? What is the probability that all the UberX’s arrive first?
Lyfts arrive first: 2!*3!/5!
Ubers arrive first: same
Q12. I write a program should print out all the numbers from 1 to 300, but prints out Fizz instead if the number is divisible by 3, Buzz instead if the number is divisible by 5, and FizzBuzz if the number is divisible by 3 and 5. What is the total number of numbers that is either Fizzed, Buzzed, or FizzBuzzed?
100+60-20=140
Q13. On a dating site, users can select 5 out of 24 adjectives to describe themselves. A match is declared between two users if they match on at least 4 adjectives. If Alice and Bob randomly pick adjectives, what is the probability that they form a match?
24C5*(1+5(24-5))/24C5*24C5 = 4/1771
Q14. A lazy high school senior types up application and envelopes to n different colleges, but puts the applications randomly into the envelopes. What is the expected number of applications that went to the right college?
1
Q15. Let’s say you have a very tall father. On average, what would you expect the height of his son to be? Taller, equal, or shorter? What if you had a very short father?
Shorter. Regression to the mean
Q16. What’s the expected number of coin flips until you get two heads in a row? What’s the expected number of coin flips until you get two tails in a row?
Q17. Let’s say we play a game where I keep flipping a coin until I get heads. If the first time I get heads is on the nth coin, then I pay you 2n-1 dollars. How much would you pay me to play this game?
less than $3
Q18. You have two coins, one of which is fair and comes up heads with a probability 1/2, and the other which is biased and comes up heads with probability 3/4. You randomly pick coin and flip it twice, and get heads both times. What is the probability that you picked the fair coin?
4/13
Data Analysis
Q19. Let’s say you’re building the recommended music engine at Spotify to recommend people music based on past lis- tening history. How would you approach this problem?
collaborative filtering
Q20. What is R2? What are some other metrics that could be better than R2 and why?
goodness of fit measure. variance explained by the regression / total variance
the more predictors you add the higher R^2 becomes.
hence use adjusted R^2 which adjusts for the degrees of freedom
or train error metrics
Q21. What is the curse of dimensionality?
High dimensionality makes clustering hard, because having lots of dimensions means that everything is “far away” from each other.
For example, to cover a fraction of the volume of the data we need to capture a very wide range for each variable as the number of variables increases
All samples are close to the edge of the sample. And this is a bad news because prediction is much more difficult near the edges of the training sample.
The sampling density decreases exponentially as p increases and hence the data becomes much more sparse without significantly more data.
We should conduct PCA to reduce dimensionality
Q22. Is more data always better?
Statistically,
It depends on the quality of your data, for example, if your data is biased, just getting more data won’t help.
It depends on your model. If your model suffers from high bias, getting more data won’t improve your test results beyond a point. You’d need to add more features, etc.
Practically,
Also there’s a tradeoff between having more data and the additional storage, computational power, memory it requires. Hence, always think about the cost of having more data.
Q23. What are advantages of plotting your data before performing analysis?
Data sets have errors. You won’t find them all but you might find some. That 212 year old man. That 9 foot tall woman.
Variables can have skewness, outliers etc. Then the arithmetic mean might not be useful. Which means the standard deviation isn’t useful.
Variables can be multimodal! If a variable is multimodal then anything based on its mean or median is going to be suspect.
Q24. How can you make sure that you don’t analyze something that ends up meaningless?
Proper exploratory data analysis.
In every data analysis task, there’s the exploratory phase where you’re just graphing things, testing things on small sets of the data, summarizing simple statistics, and getting rough ideas of what hypotheses you might want to pursue further.
Then there’s the exploitatory phase, where you look deeply into a set of hypotheses.
The exploratory phase will generate lots of possible hypotheses, and the exploitatory phase will let you really understand a few of them. Balance the two and you’ll prevent yourself from wasting time on many things that end up meaningless, although not all.
Q25. What is the role of trial and error in data analysis? What is the role of making a hypothesis before diving in?
data analysis is a repetition of setting up a new hypothesis and trying to refute the null hypothesis.
The scientific method is eminently inductive: we elaborate a hypothesis, test it and refute it or not. As a result, we come up with new hypotheses which are in turn tested and so on. This is an iterative process, as science always is.
Q26. How can you determine which features are the most important in your model?
run the features though a Gradient Boosting Machine or Random Forest to generate plots of relative importance and information gain for each feature in the ensembles.
Look at the variables added in forward variable selection
Q27. How do you deal with some of your predictors being missing?
Remove rows with missing values – This works well if 1) the values are missing randomly (see Vinay Prabhu’s answer for more details on this) 2) if you don’t lose too much of the dataset after doing so.
Build another predictive model to predict the missing values – This could be a whole project in itself, so simple techniques are usually used here.
Use a model that can incorporate missing data – Like a random forest, or any tree-based method.
Q28. You have several variables that are positively correlated with your response, and you think combining all of the variables could give you a good prediction of your response. However, you see that in the multiple linear regression, one of the weights on the predictors is negative. What could be the issue?
Multicollinearity refers to a situation in which two or more explanatory variables in a multiple regression model are highly linearly related.
Leave the model as is, despite multicollinearity. The presence of multicollinearity doesn’t affect the efficiency of extrapolating the fitted model to new data provided that the predictor variables follow the same pattern of multicollinearity in the new data as in the data on which the regression model is based.
principal component regression
Q29. Let’s say you’re given an unfeasible amount of predictors in a predictive modeling task. What are some ways to make the prediction more feasible?
PCA
Q30. Now you have a feasible amount of predictors, but you’re fairly sure that you don’t need all of them. How would you perform feature selection on the dataset?
ridge / lasso / elastic net regression
Univariate Feature Selection where a statistical test is applied to each feature individually. You retain only the best features according to the test outcome scores
“Recursive Feature Elimination”:
First, train a model with all the feature and evaluate its performance on held out data.
Then drop let say the 10% weakest features (e.g. the feature with least absolute coefficients in a linear model) and retrain on the remaining features.
Iterate until you observe a sharp drop in the predictive accuracy of the model.
Q31. Your linear regression didn’t run and communicates that there are an infinite number of best estimates for the regression coefficients. What could be wrong?
p > n.
If some of the explanatory variables are perfectly correlated (positively or negatively) then the coefficients would not be unique.
Q32. You run your regression on different subsets of your data, and that in each subset, the beta value for a certain variable varies wildly. What could be the issue here?
The dataset might be heterogeneous. In which case, it is recommended to cluster datasets into different subsets wisely, and then draw different models for different subsets. Or, use models like non parametric models (trees) which can deal with heterogeneity quite nicely.
What is the main idea behind ensemble learning? If I had many different models that predicted the same response variable, what might I want to do to incorporate all of the models? Would you expect this to perform better than an individual model or worse?
The assumption is that a group of weak learners can be combined to form a strong learner.
Hence the combined model is expected to perform better than an individual model.
Assumptions:
average out biases
reduce variance
Bagging works because some underlying learning algorithms are unstable: slightly different inputs leads to very different outputs. If you can take advantage of this instability by running multiple instances, it can be shown that the reduced instability leads to lower error. If you want to understand why, the original bagging paper( http://www.springerlink.com/cont…) has a section called “why bagging works”
Boosting works because of the focus on better defining the “decision edge”. By reweighting examples near the margin (the positive and negative examples) you get a reduced error (see http://citeseerx.ist.psu.edu/vie…)
Use the outputs of your models as inputs to a meta-model.
For example, if you’re doing binary classification, you can use all the probability outputs of your individual models as inputs to a final logistic regression (or any model, really) that can combine the probability estimates.
One very important point is to make sure that the output of your models are out-of-sample predictions. This means that the predicted value for any row in your dataframe should NOT depend on the actual value for that row.
Q33. Given that you have wi data in your o ce, how would you determine which rooms and areas are underutilized and overutilized?
If the data is more used in one room, then that one is over utilized! Maybe account for the room capacity and normalize the data.
Q34. How would you quantify the influence of a Twitter user?
like page rank with each user corresponding to the web pages and linking to the page equivalent to following.
Q35. You have 100 mathletes and 100 math problems. Each mathlete gets to choose 10 problems to solve. Given data on who got what problem correct, how would you rank the problems in terms of difficulty?
One way you could do this is by storing a “skill level” for each user and a “difficulty level” for each problem. We assume that the probability that a user solves a problem only depends on the skill of the user and the difficulty of the problem.* Then we maximize the likelihood of the data to find the hidden skill and difficulty levels.
The Rasch model for dichotomous data takes the form:
{\displaystyle \Pr\{X_{ni}=1\}={\frac {\exp({\beta _{n}}-{\delta _{i}})}{1+\exp({\beta _{n}}-{\delta _{i}})}},}
where is the ability of person and is the difficulty of item}.
Q36. You have 5000 people that rank 10 sushis in terms of salt- iness. How would you aggregate this data to estimate the true saltiness rank in each sushi?
Some people would take the mean rank of each sushi. If I wanted something simple, I would use the median, since ranks are (strictly speaking) ordinal and not interval, so adding them is a bit risque (but people do it all the time and you probably won’t be far wrong).
Q37. Given data on congressional bills and which congressio- nal representatives co-sponsored the bills, how would you determine which other representatives are most similar to yours in voting behavior? How would you evaluate who is the most liberal? Most republican? Most bipartisan?
collaborative filtering. you have your votes and we can calculate the similarity for each representatives and select the most similar representative
for liberal and republican parties, find the mean vector and find the representative closest to the center point
Q38. How would you come up with an algorithm to detect plagiarism in online content?
reduce the text to a more compact form (e.g. fingerprinting,
bag of wor
ds) then compare those with other texts by calculating the similarity
Q39. You have data on all purchases of customers at a grocery store. Describe to me how you would program an algorithm that would cluster the customers into groups. How would you determine the appropriate number of clusters include?
KNN
choose a small value of k that still has a low SSE (elbow method)
https://bl.ocks.org/rpgove/0060ff3b656618e9136b
Statistical Inference
Q40. In an A/B test, how can you check if assignment to the various buckets was truly random?
Plot the distributions of multiple features for both A and B and make sure that they have the same shape. More rigorously, we can conduct a permutation test to see if the distributions are the same.
MANOVA to compare different means
Q41. What might be the benefits of running an A/A test, where you have two buckets who are exposed to the exact same product?
Verify the sampling algorithm is random.
Q42. What would be the hazards of letting users sneak a peek at the other bucket in an A/B test?
The user might not act the same suppose had they not seen the other bucket. You are essentially adding additional variables of whether the user peeked the other bucket, which are not random across groups.
Q43. What would be some issues if blogs decide to cover one of your experimental groups?
Same as the previous question. The above problem can happen in larger scale.
Q44. How would you conduct an A/B test on an opt-in feature?
Ask someone for more details.
Q45. How would you run an A/B test for many variants, say 20 or more?
one control, 20 treatment, if the sample size for each group is big enough.
Ways to attempt to correct for this include changing your confidence level (e.g. Bonferroni Correction) or doing family-wide tests before you dive in to the individual metrics (e.g. Fisher’s Protected LSD).
Q46. How would you run an A/B test if the observations are extremely right-skewed?
lower the variability by modifying the KPI
cap values
percentile metrics
log transform
https://www.quora.com/How-would-you-run-an-A-B-test-if-the-observations-are-extremely-right-skewed
Q47. I have two different experiments that both change the sign-up button to my website. I want to test them at the same time. What kinds of things should I keep in mind?
exclusive -> ok
Q48. What is a p-value? What is the difference between type-1 and type-2 error?
type-1 error: rejecting Ho when Ho is a true
type-2 error: not rejecting Ho when Ha is true
Q49. You are AirBnB and you want to test the hypothesis that a greater number of photographs increases the chances that a buyer selects the listing. How would you test this hypothesis?
For randomly selected listings with more than 1 pictures, hide 1 random picture for group A, and show all for group B. Compare the booking rate for the two groups.
Ask someone for more details.
Q50. How would you design an experiment to determine the impact of latency on user engagement?
The best way I know to quantify the impact of performance is to isolate just that factor using a slowdown experiment, i.e., add a delay in an A/B test.
Q51. What is maximum likelihood estimation? Could there be any case where it doesn’t exist?
A method for parameter optimization (fitting a model). We choose parameters so as to maximize the likelihood function (how likely the outcome would happen given the current data and our model).
maximum likelihood estimation (MLE) is a method of estimating the parameters of a statistical model given observations, by finding the parameter values that maximize the likelihood of making the observations given the parameters. MLE can be seen as a special case of the maximum a posteriori estimation (MAP) that assumes a uniform prior distribution of the parameters, or as a variant of the MAP that ignores the prior and which therefore is unregularized.
for Gaussian mixtures, non-parametric models, it doesn’t exist
Q52. What’s the difference between a MAP, MOM, MLE estimator? In which cases would you want to use each?
MAP estimates the posterior distribution given the prior distribution and data which maximizes the likelihood function. MLE is a special case of MAP where the prior is uninformative uniform distribution.
MOM sets moment values and solves for the parameters. MOM has not used much anymore because maximum likelihood estimators have higher probability of being close to the quantities to be estimated and are more often unbiased.
Q53. What is a confidence interval and how do you interpret it?
For example, 95% confidence interval is an interval that when constructed for a set of samples each sampled in the same way, the constructed intervals include the true mean 95% of the time.
if confidence intervals are constructed using a given confidence level in an infinite number of independent experiments, the proportion of those intervals that contain the true value of the parameter will match the confidence level.
Q54. What is unbiasedness as a property of an estimator? Is this always a desirable property when performing inference? What about in data analysis or predictive modeling?
Unbiasedness means that the expectation of the estimator is equal to the population value we are estimating. This is desirable in inference because the goal is to explain the dataset as accurately as possible. However, this is not always desirable for data analysis or predictive modeling as there is the bias variance tradeoff. We sometimes want to prioritize the generalizability and avoid overfitting by reducing variance and thus increasing bias.
OTHER Important Data Science Interview Questions and Answers
Q55. What is the difference between population and sample in data?
Sample is the set of people who participated in your study whereas the population is the set of people to whom you want to generalize the results. For example – If you want to study the obesity among the children in India and you study 1000 children then those 1000 became sample whereas the all the children in the country is the population.
Sample is the subset of population.
Q56. What is the difference sample and sample frame?
Sample frame is the number of people who wanted to study whereas sample is the actual number of people who participated in your study. Ex – If you sent a marketing survey link to 300 people through email and only 100 participated in the survey then 300 is the sample survey and 100 is the sample.
Sample is the subset of sample frame. Both Sample and Sample Frame are subset of population.
Q57. What is the difference between univariate, bivariate and multivariate analysis?
Univariate analysis is performed on one variable, bivariate on two variable and multivariate analysis on two or more variables
Q58. What is difference between interpolation and extrapolation?
Extrapolation is the estimation of future values based on the observed trend on the past. Interpolation is the estimation of missing past values within two values in a sequence of values
Q59. What is precision and recall?
Precision is the percentage of correct predictions you have made and recall is the percentage of predictions that actually turned out to be true
Q60. What is confusion matrix?
Confusion matrix is a table which contains information about predicted values and actual values in a classification model
It has four parts namely true positive ,true negative, false positive and false negative
It can be used to calculate accuracy, precision and recall
Q61. What is hypothesis testing?
While performing the an experiment hypothesis testing to is used to analyze the various factors that are assumed to have an impact on the outcome of experiment
An hypothesis is some kind of assumption and hypothesis testing is used to determine whether the stated hypothesis is true or not
Initial assumption is called null hypothesis and the opposite alternate hypothesis
Q62. What is a p-value in statistics?
In hypothesis testing, p value helps to arrive at a conclusion. When p -value is too small then null hypothesis is rejected and alternate is accepted. When p-value is large then null hypothesis is accepted.
Q63. What is difference between Type-I error and Type-II error in hypothesis testing?
Type-I error is we reject the null hypothesis which was supposed to be accepted. It represents false positive
Type-II error represents we accept the null hypothesis which was supposed to be rejected. It represents false negative.
Q64. QWhat are the different types of missing value treatment?
Deletion of values
Guess the value
Average Substitution
Regression based substitution
Multiple Imputation
Q65. What is gradient descent?
When building a statistical model the objective is reduce the value of the cost function that is associated with the model. Gradient descent is an iterative optimization technique used to determine the minima of the cost function
Q66. What is difference between supervised and unsupervised learning algorithms?
Supervised learning are the class of algorithms in which model is trained by explicitly labelling the outcome. Ex. Regression, Classification
Unsupervised learning no output is given and the algorithm is made to learn the outcomes implicity Ex. Association, Clustering
Q67. What is the need for regularization in model building?
Regularization is used to penalize the model when it overfits the model. It predominantly helps in solving the overfitting problem.
Q68. Difference between bias and variance tradeoff?
High Bias is an underlying error wrong assumption that makes the model to underfit. High Variance in a model means noise in data has been too taken seriously by the model which will result in overfitting.
Typically we would like to have a model with low bias and low variance
Q69. How to solve overfitting?
Introduce Regularization
Perform Cross Validation
Reduce the number of features
Increase the number of entries
Ensembling
Q70. How will you detect the presence of overfitting?
When you build a model which has very high model accuracy on train data set and very low prediction accuracy in test data set then it is a indicator of overfitting
Q71. How do you determine the number of clusters in k-means clustering?
Elbow method ( Plotting the percentage of variance explained w.r.t to number of clusters)
Gap Statistic
Silhouette method
Q72. What is the difference between causality and correlation?
Correlation is the measure that helps us understand the relationship between two or more variables
Causation represents that causal relationship between two events. It is also known to represent cause and effect
Causation means there is correlation but correlation doesn’t necessarily mean causation
Q73. Explain normal distribution?
Normal distribution is a bell shaped curve that represents distribution of data around its mean. Any normal process would follow the normal distribution.
Most of data points tend to concentrated around the mean. If a point is further away from the mean then it is less likely to appear
Q74. What are the different ways of performing aggregation in python using pandas?
Group by function
Pivot function
Aggregate function
Q75. What are merge two list and get only unique values?
List a = [1,2,3,4] List b= [1,2,5,6] A = list(set(a+b))
Q76. How to save and retrieve model objects in python?
By using a library called pickle you can train any model and store the object in a pickle file.
When needed in future you can retrieve the object and use the model for prediction.
Q77. What is an anomaly and how is it different from outliers?
Anomaly detection is identification of items or events that didn’t fit to the exact pattern or other items in a dataset. Outliers are valid data points that are outside the norm whereas anomaly are invalid data points that are created by process that is different from process that created the other data points
Q78. What is an ensemble learning?
Ensemble learning is the art of combining more than one model to predict the final outcome of an experiment. Commonly used ensemble techniques bagging, boosting and stacking
Q79. Name few libraries that is used in python for data analysis?
Numpy
Scipy
Pandas
Scikit learn
Matplotlib\ seaborn
Q80. What are the different types of data?
Data is broadly classified into two types 1) Numerical 2) Categorical
Numerical variables is further classified into discrete and continuous data
Categorical variables
Systematic Sampling
Stratified Sampling
Quota Sampling are further classified into Binary, Nominal and Ordinal data
Q81. What is a lambda function in python?
Lambda function are used to create small, one-time anonymous function in python. It enables the programmer to create functions without a name and almost instantly
Q82. What are the different sampling methods?
Random Sampling
Systematic Sampling
Stratified Sampling
Quota Sampling
Q83. Common Data Quality Issues
Missing Values
Noise in the Data Set
Outliers
Mixture of Different Languages (like English and Chinese)
Range Constraints
Q84. What is the difference between supervised learning and un-supervised learning?
Supervised learning: Target variable is available and the algorithm learns for the train data
And applies to test data (unseen data).
Unsupervised learning: Target variable is not available and the algorithm does not need to learn
Anything beforehand.
Q85. What is Imbalanced Data Set and how to handle them? Name Few Examples?
Fraud detection
Disease screening
Imbalanced Data Set means that the population of one class is extremely large than the other
(Eg: Fraud – 99% and Non-Fraud – 1%)
Imbalanced dataset can be handled by either oversampling, undersampling and penalized Machine Learning Algorithm.
Q86. If you are dealing with 10M Data, then will you go for Machine learning (or) Deep learning Algorithm?
Machine learning algorithm suits well for small data and it might take huge amount of time to train for large data.
Whereas Deep learning algorithm takes less amount of data to train due to the help of GPU(Parallel Processing).
Q87. Examples of Supervised learning algorithm?
Linear Regression and Logistic Regression
Decision Trees and Random Forest
SVM
Naïve Bayes
XGBoost
Q88. In Logistic Regression, if you want to know the best features in your dataset then what you would do?
Apply step function, which calculates the AIC for different permutation and combination of features and provides the best features for the dataset.
Q89. What is Feature Engineering? Explain with Example?
Feature engineering is the process of using domain knowledge of the data to create features for machine learning algorithm to work
Adding more columns (or) removing columns from the existing column
Outlier Detection
Normalization etc
Q90. How to select the important features in the given data set?
In Logistic Regression, we can use step() which gives AIC score of set of features
In Decision Tree, We can use information gain(which internally uses entropy)
In Random Forest, We can use varImpPlot
Q91. When does multicollinearity problem occur and how to handle it?
It exists when 2 or more predictors are highly correlated with each other.
Example: In the Data Set if you have grades of 2nd PUC and marks of 2nd PUC, Then both gives the same trend to capture, which might internally hamper the speed and time.so we need to check if the multi collinearity exists by using VIF(variance Inflation Factor).
Note: if the Variance Inflation Factor is more than 4, then multi collinearity problem exists.
Q92. What is Variance inflation Factors (VIF)
Measure how much the variance of the estimated regression coefficients are inflated as compared to when the predictor variables are not linearly related.
Q93. Examples of Parametric machine learning algorithm and non-parametric machine learning algorithm
Parametric machine learning algorithm – Linear Regression, Logistic Regression
Non-Parametric machine learning algorithm – Decision Trees, SVM, Neural Network
Q94. What are parametric and non-parametric machine learning algorithm? And their importance
Algorithm which does not make strong assumptions are non-parametric algorithm and they are free to learn from training data.
Algorithm that makes strong assumptions are parametric and it involves 1) select the form for the function and 2)learn the coefficients for the function from training data.
Q95. When does linear and logistic regression performs better, generally?
It works better when we remove the attributes which are unrelated to the output variable and highly co-related variable to each other.
Q96. Why you call naïve bayes as “naïve” ?
Reason: It assumes that the input variable is independent, but in real world it is unrealistic, since all the features would be dependent on each other.
Q97. Give some example for false positive, false negative, true positive, true negative
False Positive – A cancer screening test comes back positive, but you don’t have cancer
False Negative – A cancer screening test comes back negative, but you have cancer
True Positive – A Cancer Screening test comes back positive, and you have cancer
True Negative – A Cancer Screening test comes back negative, and you don’t have cancer
Q98. What is Sensitivity and Specificity?
Sensitivity means “proportion of actual positives that are correctly classified” in other words “True Positive”
Specificity means “proportion of actual negatives that are correctly classified” “True Negative”
Q99. When to use Logistic Regression and when to use Linear Regression?
If you are dealing with a classification problem like (Yes/No, Fraud/Non Fraud, Sports/Music/Dance) then use Logistic Regression.
If you are dealing with continuous/discrete values, then go for Linear Regression.
Q100. What are the different imputation algorithm available?
Imputation algorithm means “replacing the Blank values by some values)
Mean imputation
Median Imputation
MICE
miss forest
Amelia
Q101. What is AIC(Akaike Information Criteria)
The analogous metric of adjusted R² in logistic regression is AIC.
AIC is the measure of fit which penalizes model for the number of model coefficients. Therefore, we always prefer model with minimum AIC value.
Q102. Suppose you have 10 samples, where 8 are positive and 2 are negative, how to calculate Entropy (important to know)
E(S) = 8/10log(8/10) – 2/10log(2/10)
Note: Log is à base 2
Q103. What is perceptron in Machine Leaning?
In Machine Learning. Perceptron is an algorithm for supervised classification of the input into one of several possible non-binary outputs
Q104. How to ensure we are not over fitting the model?
Keep the attributes/Columns which are really important
Use K-Fold cross validation techniques
Make use of drop-put incase of neural network
Q105. How the root node is predicted in Decision Tree Algorithm?
Mathematical Formula “Entropy” is utilized for predicting the root node of the tree.
Q106. What are the different Backend Process available in Keras?
TensorFlow
Theano
CNTK
Q107. Name Few Deep Learning Algorithm
TensorFlow
Theano
Lasagne
mxnet
blocks
Keras
CNTK
TFLearn
Q108. How to split the data with equal set of classes in both training and testing data?
Using Stratified Shuffle package
Q109. What do you mean by giving “epoch = 1” in neural network?
It means that “traversing the data set one time
Q110. What do you mean by Ensemble Model? When to use?
Ensemble Model is a combination of Different Models to predict correctly and with good accuracy.
Ensemble learning is used when you build component classifiers that are more accurate and independent from each other.
Q111. When will you use SVM and when to use Random Forest?
SVM can be used if the data is outlier free whereas Naïve Bayes can be used even if it has outliers (since it has built in package to take care).
SVM suits best for Text Classification Model and Random Forest suits for Binomial/Multinomial Classification Problem.
Random Forest takes care of over fitting problem with the help of tree pruning
Q112. Applications of Machine Learning?
Self Driving Cars
Image Classification
Text Classification
Search Engine
Banking, Healthcare Domain
Q113. If you are given with a use case – ‘Predict whether the transaction is fraud (or) not fraud”, which algorithm would you choose
Logistic Regression
Q114. If you are given with a use case – ‘Predict the house price range in the coming years”, which algorithm would you choose
Linear Regression
Q115. What is the underlying mathematical knowledge behind Naïve Bayes?
Bayes Theorem
Q116. When to use Random Forest and when to Use XGBoost?
If you want all core processors in your system to be utilized, then go for XGBoost(since it supports parallel processing) and if your data is small then go for random forest.
Q117. If you are training model gives 90% accuracy and test model gives 60% accuracy? Then what problem you are facing with?
Overfitting.
Overfitting and can be reduced by many methods like (Tree Pruning, Removing the minute information provided in the data set).
Q118. In Google if you type “How are “it gives you the recommendation as “How are you “/”How do you do”, this is based on what?
This kind of recommendation engine comes from collaborative filtering.
Q120. What is Boosting? Explain how Boosting works?
Boosting is a Ensemble technique that attempts to create strong classifier from a number of weak classifiers
After the first tree is created, the performance of the tree on each training instance is used to weight how much attention the next tree that is created should pay attention to each training instance by giving more weights to the misclassified one.
Models are created one after the other, each updating the weights on the training instance
Q121. What is Null Deviance and Residual Deviance (Logistic Regression Concept?)
Null Deviance indicates the response predicted by a model with nothing but an intercept
Residual deviance indicates the response predicted by a model on adding independent variables
Note:
Lower the value, better the model
Q122. What are the different method to split the tree in decision tree?
Information gain and gini index
Q123. What is the weakness for Decision Tree Algorithm?
Not suitable for continuous/Discrete variable
Performs poorly on small data
Q124. Why do we use PCA(Principal Components Analysis) ?
These are important feature extraction techniques used for dimensionality reduction.
Q125. During Imbalanced Data Set, will you
Calculate the Accuracy only? (or)
Precision, Recall, F1 Score separately
We need to calculate precision, Recall separately
Q126. How to ensure we are not over fitting the model?
Keep the attributes/Columns which are really important
Use K-Fold cross validation techniques
make use of drop-put in case of neural network
Q127. Steps involved in Decision Tree and finding the root node for the tree
Step 1:- How to find the Root Node
Use Information gain to understand the each attribute information w.r.t target variable and place the attribute with the highest information gain as root node.
Step 2:- How to Find the Information Gain
Please apply the entropy (Mathematical Formulae) to calculate Information Gain. Gain (T,X) = Entropy(T) – Entropy(T,X) here represent target variable and X represent features.
Step3: Identification of Terminal Node
Based on the information gain value obtained from the above steps, identify the second most highest information gain and place it as the terminal node.
Step 4: Predicted Outcome
Recursively iterate the step4 till we obtain the leaf node which would be our predicted target variable.
Step 5: Tree Pruning and optimization for good results
It helps to reduce the size of decision trees by removing sections of the tree to avoid over fitting.
Q128. What is hyper plane in SVM?
It is a line that splits the input variable space and it is selected to best separate the points in the input variable space by their class(0/1,yes/no).
Q129. Explain Bigram with an Example?
Eg: I Love Data Science
Bigram – (I Love) (Love Data) (Data Science)
Q130. What are the different activation functions in neural network?
Relu, Leaky Relu , Softmax, Sigmoid
Q131. Which Algorithm Suits for Text Classification Problem?
SVM, Naïve Bayes, Keras, Theano, CNTK, TFLearn(Tensorflow)
Q132. You are given a train data set having lot of columns and rows. How do you reduce the dimension of this data?
Principal Component Analysis(PCA) would help us here which can explain the maximum variance in the data set.
We can also check the co-relation for numerical data and remove the problem of multi-collinearity(if exists) and remove some of the columns which may not impact the model.
We can create multiple dataset and execute them batch wise.
Q133. You are given a data set on fraud detection. Classification model achieved accuracy of 95%.Is it good?
Accuracy of 96% is good. But we may have to check the following items:
what was the dataset for the classification problem
Is Sensitivity and Specificity are acceptable
if there are only less negative cases, and all negative cases are not correctly classified, then it might be a problem
In-Addition it is related to fraud detection, hence needs to be careful here in prediction (i.e not wrongly predicting the fraud as non-fraud patient.
Q134. What is prior probability and likelihood?
Prior probability:
The proportion of dependent variable in the data set.
Likelihood:
It is the probability of classifying a given observation as ‘1’ in the presence of some other variable.
Q135. How can we know if your data is suffering from low bias and high variance?
Random Forest Algorithm can be used to tackle high variance problem.in the cases of low bias and high variance L1,L2 regularization can help.
Q134. How is kNN different from kmeans clustering?
Kmeans partitions a data set into clusters, which is homogeneous and points in the cluster are close to each other. Whereas KNN tries to classify unlabelled observation based on its K surrounding neighbours.
Q135. Random Forest has 1000 trees, Training error: 0.0 and validation error is 20.00.What is the issue here?
It is the classical example of over fitting. It is not performing well on the unseen data. We may have to tune our model using cross validation and other techniques to overcome over fitting
Q136. Data set consisting of variables having more than 30% missing values? How will you deal with them?
We can remove them, if it does not impact our model
We can apply imputation techniques (like MICE, MISSFOREST,AMELIA) to avoid missing values
Q137. What do you understand by Type I vs. Type II error?
Type I error occurs when – “we classify a value as positive, when the actual value is negative”
(False Positive)
Type II error occurs when – “we classify a value as negative, when the actual value if positive”
(False Negative)
Q138. Based on the dataset, how will you know which algorithm to apply ?
If it is classification related problem,then we can use logistic,decision trees etc…
If it is Regression related problem, then we can use Linear Regression.
If it is Clustering based, we can use KNN.
We can also apply XGB, RF for better accuracy.
Q139. Why normalization is important?
Data Set can have one column in the range (10,000/20,000) and other column might have data in the range (1, 2, 3).clearly these two columns are in different range and cannot accurately analyse the trend. So we can apply normalization here by using min-max normalization (i.e to convert it into 0-1 scale).
https://mytecbooks.blogspot.com/2019/06/data-science-experienced-level.html