A machine learning engineer is working on a customer churn prediction project. The dataset contains 100,000 records with 15 features, including customer demographics, account information, and usage patterns. The target variable 'churned' is binary with 15% positive examples. During EDA, the engineer notices that the feature 'tenure' (number of months the customer has been with the company) has a multimodal distribution with peaks at 1, 12, 24, and 36 months. Also, the feature 'monthly_charges' has a strong positive correlation with 'total_charges' (correlation coefficient = 0.95). The engineer wants to build a logistic regression model. Which preprocessing steps should the engineer take to address these issues? (Select TWO.)
Trap 1: Apply log transformation to the 'tenure' feature to make it…
Log transformation does not make a multimodal distribution unimodal; it only compresses the scale.
Trap 2: Create polynomial features up to degree 3 for 'tenure' to capture…
Polynomial features can introduce multicollinearity and are not the best approach for multimodal distributions; binning is more interpretable.
Trap 3: Standardize all numerical features to have mean 0 and variance 1.
Standardization is good for logistic regression with regularization, but it does not address the specific issues of multicollinearity and multimodality mentioned.
- A
Bin the 'tenure' feature into categorical groups (e.g., 0-6, 7-12, 13-24, 25-36, 36+) to capture the non-linear relationship.
Binning can effectively capture the peaks in the distribution and model the non-linear effect of tenure on churn.
- B
Remove one of the correlated features, such as 'total_charges', to reduce multicollinearity.
Highly correlated features can cause instability in logistic regression coefficient estimates; removing one helps.
- C
Apply log transformation to the 'tenure' feature to make it unimodal.
Why wrong: Log transformation does not make a multimodal distribution unimodal; it only compresses the scale.
- D
Create polynomial features up to degree 3 for 'tenure' to capture non-linearity.
Why wrong: Polynomial features can introduce multicollinearity and are not the best approach for multimodal distributions; binning is more interpretable.
- E
Standardize all numerical features to have mean 0 and variance 1.
Why wrong: Standardization is good for logistic regression with regularization, but it does not address the specific issues of multicollinearity and multimodality mentioned.