Exam AWS Certified Machine Learning - Specialty topic 1 question 217 discussion

oversample the minority class

Undersampling not an option for already limited observations. SMOTE clearly MOST promising first action before trying to balance classes

Do we need to stratify the non-failure events by machine type?

B. Oversample the failure cases by using the Synthetic Minority Oversampling Technique (SMOTE). Since the number of failure cases is relatively small, oversampling the failure cases using techniques like SMOTE can help balance the class distribution and prevent the model from being biased towards the majority class. SMOTE creates synthetic samples for the minority class by interpolating new samples between existing ones. This will help improve the model's accuracy in predicting failure cases. Adjusting class weights (A) or undersampling (C, D) may not be as effective in this scenario.

The data provided is imbalanced, with only 100 failure cases out of 10,000 event samples. Therefore, it is important to address this imbalance to improve the accuracy of the predictive maintenance model.