Supervised Learning

Question 1

What is Reinforcement Learning?

Answer 1

Software agent optimizes its behaviour based on rewards and punishments.

Question 2

How to do VEDA for 2 categorical variables?

Answer 2

Using sns.barplot(x=”day”, y=”total_bill”, data=tips) where data is a df.
OR tips.boxplot(‘day’, ‘total_bill’)

Question 3

How to do VEDA for binary categorical variables?

Answer 3

plt. figure()
sns. countplot(x=’education’, hue=’party’, data=df, palette=’RdBu’)
plt. xticks([0,1], [‘No’, ‘Yes’])
plt. show()

Question 4

How to do pair-wise VEDA for 4 quantitative variables?

Answer 4

pd.scatter_matrix(df, c = y, figsize = [8, 8], marker = ‘D’)

Question 5

What is accuracy?

Answer 5

Fraction of correct predictions.

Question 6

How to access predictor values after removing target values from df?

Answer 6

df.drop(‘target’, axis=1).values

Question 7

How to turn list of values into format for sklearn?

Answer 7

X.reshape(-1,1)

Question 8

How to generate pairwise feature correlation VEDA?

Answer 8

sns.heatmap(df.corr(), square=True, cmap=’RdYlGn’)

Question 9

What are a and b in y = ax +b ?

Answer 9

a is slope and b is y intercept

Question 10

How to do k-fold cv with sklearn?

Answer 10

cross_val_score(reg, X, y, cv=k)

Question 11

Why should regularization be used?

Answer 11

To penalize large coefficients and avoid over-fitting

Question 12

What is Ridge regression?

Answer 12

regression with regularization where alpha (hyper)-parameter weighs the OLS. Should be first choice for regression over lasso.

Question 13

What is Lasso regression?

Answer 13

regression with regularization where coefs can be set to 0 to remove unimportant features. Great for feature selection.

Question 14

How to specify parameters for Lasso and access its coefficients?

Answer 14

lasso = Lasso(alpha=0.4, normalize=True)

lasso.coef_

Question 15

When is accuracy a poor metric when only fraction of correct predictions is used?

Answer 15

When there is class imbalance where low freq items will never be correctly labeled. if maj class is 99%, then accuracy of 99% can be achieved with model that always picks maj class.

Question 16

How to calculate accuracy using confusion matrix?

Answer 16

(TP+TN)/(TP+TN+FN+FP)

Question 17

How to calculate precision using conf matrix?

Answer 17

TP/TP+FP

Question 18

How to calculate recall using conf matrix?

Answer 18

TP/TP+FN

Question 19

What is F1?

Answer 19

(2(precision*recall))/(precision+recall)

Question 20

How to get classifier’s performance report using SKL?

Answer 20

sklearn.metrics.classification_report

Question 21

What is the ROC curve?

Answer 21

The receiver operating characteristic curve is a graphical plot that illustrates the diagnostic ability of a binary classifier system as its discrimination threshold is varied.

Question 22

What is a better than random AUC?

Answer 22

If the AUC is greater than 0.5, the model is better than random guessing.

Question 23

How to use AUC during cross-validation?

Answer 23

cross_val_score(logreg, X, y, cv=5, scoring=’roc_auc’)

Question 24

What are hyperparameters?

Answer 24

Parameters that cannot be learnt by fitting the model. Like k or alpha.

Question 25

What is C in LogReg?

Answer 25

C controls the inverse of the regularization strength. A large C can lead to an overfit model, while a small C can lead to an underfit model.

Question 26

How to grid search with SKL?

Answer 26

GridSearchCV(logreg, {'C': c_space}, cv=5)

Question 27

What to do since scikit-learn does not accept non-numerical features?

Answer 27

Use one-hot encoding using sklearn or pandas.get_dummies(df)

Question 28

What is the step used to avoid dropping nans?

Answer 28

Inputing data by making educated guess on what it should be, such as mean or mode: imp = Imputer(missing_values='NaN', strategy='most_frequent', axis=0)

Question 29

Where to find scaler in SKL?

Answer 29

from sklearn.preprocessing import StandardScaler

Question 30

How to configure pipeline with scaler and k-nn clf?

Answer 30

pipeline([('scaler', StandardScaler()), | ('knn', KNeighborsClassifier())])

Question 31

What is gamma in SVM?

Answer 31

gammagamma controls the kernel coefficient

Question 32

What does axis=1 mean in Pandas?

Answer 32

It means across columns (to the right)

Question 33

What is log loss?

Answer 33

``` The metric (often used for multi-class classification) is negative the log likelihood of the model that says each test observation is chosen independently from a distribution that places the submitted probability mass on the corresponding class, for each observation. Log loss provides a steep penalty for predictions that are both wrong and confident. ```

Question 34

How to see source of python function?

Answer 34

inspect.getsource

Question 35

How to keep data distribution when using train/test split?

Answer 35

Stratification

Question 36

How to limit size of text features?

Answer 36

Using hashing trick to perform dimensionality reduction.

Question 37

How to add feature interaction in a model?

Answer 37

Using polynomials

Question 38

How to change scale of plot axis?

Answer 38

plt.xscale('log')