AWS Certified Machine Learning – SpecialtyModelingMedium
A data science team is developing a fraud detection model. The dataset is severely imbalanced, with fraudulent transactions representing less than 0.1% of the total. The initial model, trained on the raw dataset, shows high accuracy but very low recall for fraudulent transactions. To improve the model's ability to detect fraud, the team decides to synthetically generate new samples for the minority class (fraudulent transactions) to balance the dataset. Which technique is commonly used for this purpose?
- APrincipal Component Analysis (PCA)
- BOversampling
- CSMOTE (Synthetic Minority Over-sampling Technique)
- DUndersampling
Show answer & explanationAnswer & explanation
Correct answer: C. SMOTE (Synthetic Minority Over-sampling Technique)
SMOTE (Synthetic Minority Over-sampling Technique) is specifically designed to address class imbalance by creating synthetic samples for the minority class. It works by selecting a minority class instance and its k-nearest neighbors, then generating new synthetic instances along the line segments connecting them.
Why the other options are wrong
- A. PCA is a dimensionality reduction technique and does not address class imbalance by generating synthetic samples.
- B. Oversampling generally refers to simply duplicating existing minority class samples, which can lead to overfitting without introducing new information. SMOTE is a more sophisticated form of oversampling.
- D. Undersampling reduces the number of samples in the majority class, which can lead to loss of valuable information and is not about generating new minority samples.
SMOTE
Synthetic Minority Over-sampling Technique, an oversampling method that generates synthetic samples for the minority class by interpolating between existing minority class instances and their nearest neighbors.
- Addresses class imbalance by increasing minority class size.
- Creates synthetic, not duplicate, samples.
- Helps prevent overfitting compared to simple oversampling.
Memory trick: SMOTE's the antidote, for tiny classes, it's the vote, creating new friends, so the model can quote.