MLA-C01 Question 59
Select 3You are working on a machine learning model to predict customer churn for a subscription service. The dataset includes features such as 'MonthlyCharges', 'TotalCharges', and 'Tenure'. The 'MonthlyCharges' and 'TotalCharges' features have wide ranges and different units. The 'Tenure' feature represents the number of months a customer has been subscribed to the service. What feature engineering techniques should you apply to improve the model's performance?
- A
Normalize the 'MonthlyCharges' and 'TotalCharges' features to bring their values into a common scale.
- B
Apply log transformation to the 'Tenure' feature to handle non-linear relationships.
- C
Split the 'TotalCharges' feature into bins based on value ranges, such as low, medium, and high spenders.
- D
Standardize the 'MonthlyCharges' and 'TotalCharges' features to have a mean of 0 and a standard deviation of 1.
- E
Apply one-hot encoding to the 'Tenure' feature to convert it into categorical data.
Show answer and explanation
Correct answers: A, C, D
Explanation
The features 'MonthlyCharges' and 'TotalCharges' have wide ranges and different units, which could negatively impact the model's performance. Normalizing or standardizing these features ensures they are on a comparable scale and prevents bias in algorithms sensitive to feature magnitudes. Additionally, binning 'TotalCharges' can help capture non-linear relationships, especially in tree-based models. However, log transformation and one-hot encoding are not applicable in this scenario as they do not align with the data characteristics or the type of features provided.
- A. Correct.
Normalization is a valid technique to scale features like 'MonthlyCharges' and 'TotalCharges' to a common range, making them more comparable and improving model performance.
- B. Incorrect.
A log transformation may not be suitable for the 'Tenure' feature, as it is already linear and does not exhibit a skewed distribution or exponential growth that would benefit from a log transformation.
- C. Correct.
Binning 'TotalCharges' into categories such as low, medium, and high spenders can help capture non-linear relationships and simplify the feature for certain models.
- D. Correct.
Standardizing features like 'MonthlyCharges' and 'TotalCharges' to have a mean of 0 and a standard deviation of 1 can improve performance for algorithms sensitive to feature scaling, such as support vector machines and gradient descent-based models.
- E. Incorrect.
One-hot encoding is not appropriate for the 'Tenure' feature since it is a continuous numerical variable, not a categorical one.