Navigating Bias-Variance and Imbalanced Classes in Fraud Detection

Imagine you're tasked with developing a classification model to identify fraudulent transactions for a credit card company. You have access to a decade's worth of transaction data, each marked as either fraudulent or legitimate. Outline your approach to constructing a fraud detection model, considering various model types, addressing the bias-variance tradeoff, and tackling challenges posed by class imbalance. Available input features include: - Transaction amount - Merchant category - Merchant zip code - Billing address zip code - Average transaction amount for the past six months Output feature: - Fraud indicator (0 = legitimate, 1 = fraudulent) Note: Fraudulent transactions account for a mere 0.01% of the dataset. As you design the model, remember to balance the bias-variance tradeoff and address the significant class imbalance.