Build a Payment Fraud Detection Model
Company: Adyen
Role: Machine Learning Engineer
Category: Machine Learning
Difficulty: easy
Interview Round: HR Screen
You are interviewing for a Machine Learning Engineer role at a FinTech company.
Part 1: Explain the following ML fundamentals:
- What is overfitting?
- How can you detect overfitting during model development?
- How do L1 and L2 regularization reduce overfitting, and how do they differ?
Part 2: You are given a payment-transaction dataset for fraud detection. Each row represents one transaction and includes a binary label `is_fraud`, along with typical transaction features such as amount, timestamp-derived features, merchant category, country, payment method, device attributes, and customer history aggregates.
Build a runnable machine-learning pipeline that trains a fraud detection model. Your solution should:
- Split the data into train, validation, and test sets without leaking future information.
- Handle missing values and categorical features.
- Address severe class imbalance.
- Train at least one reasonable baseline model.
- Evaluate the model using metrics appropriate for fraud detection.
- Explain what you would improve if you had more time.
You may use standard ML libraries and AI coding tools, but the final code must run end-to-end.
Quick Answer: This question evaluates machine-learning fundamentals (overfitting, L1/L2 regularization), data preprocessing and engineering (temporal train/validation/test splitting, missing values, categorical features), class-imbalance mitigation, model selection and evaluation within the payment fraud detection domain.