MLS-C01 Question 141
Select 3A financial services company wants to predict whether a transaction is fraudulent or not. The dataset contains millions of labeled transactions with a high class imbalance, where fraudulent transactions make up only 1% of the data. The solution must ensure accurate predictions for the minority class and be scalable to handle the large dataset. Which model(s) would be most appropriate for this scenario?
- A
XGBoost with class weighting
- B
Logistic Regression without any modifications
- C
Random Forest with SMOTE (Synthetic Minority Oversampling Technique)
- D
Deep Neural Network with no adjustments for class imbalance
- E
Support Vector Machine (SVM) with a radial basis function (RBF) kernel and class weighting
Show answer and explanation
Correct answers: A, C, E
Explanation
This scenario requires a model that can handle a highly imbalanced dataset while scaling to millions of transactions. XGBoost with class weighting, Random Forest with SMOTE, and SVM with class weighting are all effective approaches for addressing class imbalance and ensuring accurate predictions for the minority class. Logistic Regression without modifications and Deep Neural Networks without adjustments are not suitable as they fail to address the class imbalance, which would compromise the model's performance on fraudulent transactions.
- A. Correct.
XGBoost with class weighting is a good choice because it can handle class imbalance effectively by assigning higher weights to the minority class during training. Additionally, XGBoost is highly scalable and efficient for large datasets, making it suitable for this use case.
- B. Incorrect.
Logistic Regression without any modifications is not appropriate because it does not address the class imbalance issue, which would result in poor performance for the minority class (fraudulent transactions).
- C. Correct.
Random Forest with SMOTE is a valid option because SMOTE can generate synthetic samples for the minority class, addressing the class imbalance. Random Forest is also effective for handling large datasets.
- D. Incorrect.
Deep Neural Network with no adjustments for class imbalance is not suitable in this scenario because it would likely overfit to the majority class without specific techniques to handle the imbalance, such as cost-sensitive training or oversampling.
- E. Correct.
Support Vector Machine (SVM) with an RBF kernel and class weighting is appropriate as it can handle class imbalance by adjusting class weights, which helps the model focus more on the minority class. However, SVMs may struggle with extremely large datasets unless optimized.