Answer-first summary for fast verification
Answer: Augment the dataset by increasing the number of fraudulent transactions tenfold through oversampling., Apply Synthetic Minority Over-sampling Technique (SMOTE) to generate synthetic samples of the fraudulent transactions.
Oversampling the fraudulent transactions (Option D) and applying SMOTE (Option E) are both effective strategies for addressing the class imbalance in the dataset, thereby enhancing the model's ability to detect fraudulent transactions. Oversampling increases the representation of fraudulent transactions, while SMOTE generates synthetic samples to provide a more balanced dataset. Both methods help in reducing the cost of false negatives and comply with the bank's regulatory requirements by improving the model's accuracy in fraud detection.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
In the context of developing a random forest model for fraud detection at a bank, the dataset comprises transactions with only 1% identified as fraudulent. The bank is concerned about the high cost of false negatives and requires a solution that not only improves the model's ability to detect fraudulent transactions but also adheres to strict compliance regulations. Which of the following data transformation strategies would most effectively enhance the classifier's performance under these constraints? (Choose two options)
A
Convert your data into TFRecords format for efficient storage.
B
Implement one-hot encoding across all categorical features to ensure compatibility.
C
Standardize all numerical features by applying z-score normalization.
D
Augment the dataset by increasing the number of fraudulent transactions tenfold through oversampling.
E
Apply Synthetic Minority Over-sampling Technique (SMOTE) to generate synthetic samples of the fraudulent transactions.
No comments yet.