NCA-GENL Question 33
Select 3You are tasked with building a machine learning model to predict customer churn for a subscription service. The dataset contains features such as 'Monthly Payment', 'Contract Length', 'Customer Tenure', and 'Support Tickets Raised'. To ensure the model generalizes well to unseen data, which of the following steps should you include in your workflow?
- A
Perform cross-validation during model training to evaluate performance on unseen data.
- B
Standardize numerical features like 'Monthly Payment' and 'Customer Tenure' before training.
- C
Train the model exclusively on the entire dataset to maximize its accuracy.
- D
Compare multiple models (e.g., logistic regression, decision trees) to select the best-performing one.
- E
Remove all features with missing values, regardless of their importance to the prediction task.
Show answer and explanation
Correct answers: A, B, D
Explanation
To build a robust machine learning model, it is essential to use cross-validation to evaluate generalization performance, standardize numerical features to ensure consistent scaling, and compare multiple models to identify the best one. Avoid training on the entire dataset without validation and be cautious when handling missing data to preserve important features.
- A. Correct.
Correct: Cross-validation is a standard technique to evaluate a model's performance on unseen data by splitting the data into training and validation sets multiple times.
- B. Correct.
Correct: Standardizing numerical features helps ensure consistent scaling, which is particularly important for models sensitive to feature magnitudes, such as support vector machines or gradient descent optimizers.
- C. Incorrect.
Incorrect: Training on the entire dataset without reserving a validation or test set will likely lead to overfitting, as the model will not be evaluated on unseen data.
- D. Correct.
Correct: Comparing multiple models is a key step in machine learning to identify the one that performs best for the given dataset and task.
- E. Incorrect.
Incorrect: Blindly removing features with missing values can lead to the loss of valuable information. Instead, techniques like imputation should be considered.