MLA-C01 Question 57
Select 3You are building a machine learning model to predict housing prices. The dataset contains features such as square footage, the year the house was built, and the number of bedrooms. The square footage feature has a wide range of values, from 500 to 10,000. Your preliminary tests show that the model's performance is poor. Which feature engineering techniques should you apply to improve the model's performance?
- A
Apply min-max normalization to the square footage feature
- B
Perform log transformation on the square footage feature
- C
Split the square footage feature into bins (e.g., small, medium, large)
- D
Remove the square footage feature as it has a wide range
- E
Standardize the square footage feature by subtracting the mean and dividing by the standard deviation
Show answer and explanation
Correct answers: A, B, E
Explanation
Feature engineering is crucial for improving model performance. For the square footage feature, techniques like min-max normalization, log transformation, and standardization effectively handle its wide range of values without losing information. These methods ensure that the feature is properly scaled and normalized, improving the model's ability to learn patterns from the data. Removing the feature or using binning would either decrease model performance or result in information loss, making them less suitable choices in this case.
- A. Correct.
Min-max normalization scales the feature values to a range of 0 to 1, which can help models converge faster during training. This is particularly effective for features with a wide range of values like square footage.
- B. Correct.
Log transformation can reduce the impact of outliers and compress a wide range of values into a smaller range, which benefits models that are sensitive to large numerical variations.
- C. Incorrect.
Binning reduces feature granularity but can lead to loss of information. While it might be useful in some scenarios, it is not optimal for a regression problem like predicting housing prices.
- D. Incorrect.
Removing the square footage feature would discard valuable information. Square footage is highly relevant to predicting housing prices, so dropping it would likely degrade model performance.
- E. Correct.
Standardization (z-score scaling) brings the feature to a mean of 0 and a standard deviation of 1, which is beneficial for models that assume normally distributed features, such as linear regression or SVMs. It is an effective way to handle features with a wide range of values.