MLA-C01 Question 2
Select 3You are building a machine learning model to predict customer churn for a subscription-based service. The dataset contains missing values, categorical variables, and features with varying scales. Additionally, the dataset contains a 'customer_id' field which uniquely identifies each customer. As part of the data preparation process, which of the following steps should you perform to ensure the data is ready for training? (Choose TWO)
- A
Remove the 'customer_id' field from the dataset before training the model.
- B
Scale numerical features to ensure they are on a similar range.
- C
Fill missing values in the dataset using a unique identifier for each row.
- D
Convert categorical variables into numerical representations (e.g., one-hot encoding or label encoding).
- E
Normalize the target variable to fall within the range [0, 1].
Show answer and explanation
Correct answers: A, B, D
Explanation
Preparing data for machine learning involves addressing issues like irrelevant features, feature scaling, handling missing values, and encoding categorical variables. Removing identifiers like 'customer_id' prevents noise, scaling ensures all features contribute equally, and encoding categorical variables makes them suitable for model training. Missteps like inappropriate imputation methods or unnecessary normalization of the target variable should be avoided.
- A. Correct.
Correct: The 'customer_id' field is not a feature and does not contribute to the prediction task. Keeping it in the dataset can introduce noise or bias, so it must be removed before training.
- B. Correct.
Correct: Scaling numerical features helps algorithms (e.g., gradient-based models) perform better by ensuring all features contribute equally to the model's learning process.
- C. Incorrect.
Incorrect: Filling missing values with a unique identifier for each row is not a valid imputation technique. Missing values should be handled using statistical methods (e.g., mean, median), domain knowledge, or other appropriate strategies.
- D. Correct.
Correct: Machine learning models require numerical inputs. Categorical variables need to be converted into numerical representations for the model to process them effectively.
- E. Incorrect.
Incorrect: Normalizing the target variable depends on the problem and algorithm. It is not a required step for data preparation unless the model or problem dictates it.