Algorithms

Question 1

algorithm for binary classification, multiple features, not enough data for a neural net

Answer 1

binary logistic regression

Question 2

algorithm for predicting a ranking (number of stars given to a movie or something like that)

Answer 2

ordinal logistic regression

Question 3

image classification algorithm?

Answer 3

CNN

Question 4

time series forecasting algorithms?

Answer 4

bayesian structural time series models, LSTMs, RNNs, ARIMA, SARIMA, SARIMAX

Question 5

5 assumptions of linear regression?

Answer 5

1) linear relation between features and target
2) little or no multicollinearity between features
3) homeoskedasticity
4) residuals (uncertainty) follow a normal distribution
5) Little or no autocorrelation in residuals (usually happens in time series where one step is dependent on the last).

Question 6

If we have two correlated features, what do we do?

Answer 6

Combine them to make an independent feature or drop one

Question 7

Dimensionality reduction?

Answer 7

PCA

Question 8

What metric is best for linear regression and why?

Answer 8

R^2; for linear regression, it represents (explained variance)/(total variance)

Question 9

How does a decision tree decide what split to make at each node?

Answer 9

For classification: minimize Gini impurity

For regression: