Databricks Machine Learning Associate Question 551
Select 4You are training a machine learning model on a dataset where one class significantly outnumbers the others, causing the model to perform poorly on the minority class. Which of the following techniques can help mitigate the effects of this data imbalance?
- A
Oversample the minority class using techniques like SMOTE (Synthetic Minority Oversampling Technique)
- B
Undersample the majority class to reduce its dominance in the dataset
- C
Use a weighted loss function to penalize the model more for errors on the minority class
- D
Normalize all feature values to bring them within the same range
- E
Collect more data for the minority class to improve its representation
Show answer and explanation
Correct answers: A, B, C, E
Explanation
Addressing data imbalance is crucial for training effective machine learning models, especially when working with imbalanced datasets. Techniques like oversampling, undersampling, using weighted loss functions, and collecting additional data for the minority class are all effective mitigation strategies. However, normalization is unrelated to the issue of data imbalance and focuses on feature scaling instead.
- A. Correct.
Oversampling the minority class generates synthetic examples or duplicates existing ones, helping to balance the dataset. SMOTE is a common technique for this purpose.
- B. Correct.
Undersampling the majority class reduces its representation, making the dataset more balanced, though it may risk losing valuable information.
- C. Correct.
Using a weighted loss function ensures that the model focuses more on correctly predicting the minority class, mitigating imbalance-induced bias.
- D. Incorrect.
Normalizing feature values is a preprocessing step for ensuring consistent scaling but does not address data imbalance directly.
- E. Correct.
Collecting more data for the minority class improves its representation and reduces the imbalance, though it may not always be feasible.