Data Scientist Interview QnA
Company: Genpact
1. Inter quartile ranges?
The most effective way to find all of your outliers is by using the interquartile range (IQR). The IQR contains the middle bulk of your data, so outliers can be easily found once you know the IQR. Quartiles divide the entire set into four equal parts. So, there are three quartiles, first, second and third represented by Q 1 , Q 2 and Q 3 , respectively. Q 2 is nothing but the median.
2. Imputation methods?
They are:
List wise or case deletion
Pairwise deletion
Mean substitution
Regression imputation
Maximum likelihood.
3. Deal overfittting?
Techniques to reduce overfitting:
Increase training data.
Reduce model complexity.
Early stopping during the training phase.
Ridge Regularization and Lasso Regularization.
4. Gridsearch vs Random search?
Random search differs from grid search in that we no longer provide an explicit set of possible values for each hyperparameter; rather, we provide a statistical distribution for each hyperparameter from which values are sampled. Essentially, we define a sampling distribution for each hyperparameter to carry out a randomized search.
5. Hyperparameters in SVM?
kernel: It maps the observations into some feature space. Ideally the observations are more easily (linearly) separable after this transformation. There are multiple standard kernels for this transformations, e.g. the linear kernel, the polynomial kernel and the radial kernel.
C: It is a hypermeter in SVM to control error. The C parameter tells the SVM optimization how much you want to avoid misclassifying each training example. For large values of C, the optimization will choose a smaller-margin hyperplane if that hyperplane does a better job of getting all the training points classified correctly.
gamma: Gamma is used when we use the Gaussian RBF kernel. if you use linear or polynomial kernel then you do not need gamma only you need C hypermeter. Somewhere it is also used as sigma. Gamma decides that how much curvature we want in a decision boundary.
6. Ridge vs lasso?
Ridge and Lasso regression are some of the simple techniques to reduce model complexity and prevent over-fitting which may result from simple linear regression . Ridge Regression : In ridge regression, the cost function is altered by adding a penalty equivalent to square of the magnitude of the coefficients. lasso (least absolute shrinkage and selection operator; also Lasso or LASSO) is a regression analysis method that performs both variable selection and regularization in order to enhance the prediction accuracy and interpretability of the resulting statistical model.
