MLA-C01 Question 63
Single answerYou are building a machine learning model to predict customer churn for a telecommunications company. The dataset contains a categorical feature named 'Contract Type,' which includes values such as 'Month-to-Month,' 'One Year,' and 'Two Year.' The machine learning algorithm you are using does not natively support categorical data. Which encoding technique would be most appropriate if you are concerned about the model's interpretability and want to avoid introducing unnecessary complexity?
- A
One-hot encoding
- B
Label encoding
- C
Binary encoding
- D
Tokenization
Show answer and explanation
Correct answer: A
Explanation
One-hot encoding is the most appropriate technique in this scenario because it avoids introducing ordinal relationships between categories, which is critical when the categories do not have a natural order (e.g., 'Month-to-Month,' 'One Year,' 'Two Year'). This technique also enhances interpretability, as each category is represented by its own binary column. While label encoding and binary encoding are valid techniques, they are less suitable here due to their limitations regarding interpretability and the potential for misrepresenting the relationships between categories.
- A. Correct.
One-hot encoding is appropriate because it converts each category into a separate binary column. This approach is interpretable as it avoids ordinal relationships between categories but can increase dimensionality if there are many unique categories.
- B. Incorrect.
Label encoding assigns a unique integer value to each category. However, it introduces an ordinal relationship between categories, which might not be suitable depending on the algorithm. It is less interpretable compared to one-hot encoding.
- C. Incorrect.
Binary encoding is a compact representation of categorical values using binary digits. While it reduces dimensionality, it can be harder to interpret compared to one-hot encoding and is not necessary in this scenario where the number of categories is small.
- D. Incorrect.
Tokenization is used for text data and is not a relevant encoding technique for the given categorical feature in this scenario.