MLS-C01 Question 103
Select 4You are tasked with building a machine learning model to predict customer churn for a subscription-based business. The dataset includes categorical features such as 'Subscription Plan' (Basic, Standard, Premium), numerical features such as 'Monthly Usage Hours', and outliers in the 'Lifetime Value' column. Additionally, the dataset contains high-dimensional features like 'Customer Activity Logs' with 500 columns. Which actions should you take to prepare the dataset effectively for the model?
- A
Apply one-hot encoding to the 'Subscription Plan' feature.
- B
Use binning to group customers in 'Lifetime Value' into ranges such as Low, Medium, and High.
- C
Remove the 'Customer Activity Logs' feature entirely to reduce dimensionality.
- D
Apply Principal Component Analysis (PCA) to the 'Customer Activity Logs' feature to reduce dimensionality.
- E
Scale the 'Monthly Usage Hours' feature using standardization or normalization.
Show answer and explanation
Correct answers: A, B, D, E
Explanation
This question tests your ability to analyze and evaluate feature engineering concepts. Preparing the dataset involves encoding categorical variables, handling outliers, reducing the dimensionality of high-dimensional features, and scaling numerical features. These steps ensure the data is in a format suitable for machine learning algorithms while preserving important information for the model to learn effectively.
- A. Correct.
One-hot encoding is appropriate for categorical features like 'Subscription Plan', as it allows the model to handle these non-numeric values effectively.
- B. Correct.
Binning can help reduce the impact of outliers in the 'Lifetime Value' feature by grouping data into more interpretable ranges, which can improve model performance.
- C. Incorrect.
Removing the 'Customer Activity Logs' feature entirely may lead to a loss of valuable information. Dimensionality reduction techniques like PCA are better suited for high-dimensional features.
- D. Correct.
Principal Component Analysis (PCA) is an effective technique to reduce the dimensionality of the 'Customer Activity Logs' feature while retaining important information.
- E. Correct.
Scaling numerical features such as 'Monthly Usage Hours' ensures that the model treats them appropriately, especially for algorithms sensitive to feature scaling like gradient descent.