Credit Score Classifier

The Problem

A bank needs to decide how much credit to offer new customers, and whether to offer credit at all. Instead of manually reviewing each customer's financial history, this project builds a model that predicts a customer's credit score category based on their financial profile, so credit decisions can be automated and scaled.

Approach

1. Loaded historical data for 100,000 customers, including salary, number of bank accounts/credit cards, loan interest rates, payment delays, credit mix, total debt, credit utilization, and monthly investment behavior.
2. Encoded categorical fields (profession, credit mix, payment behavior) into numeric values using Label Encoding.
3. Split the data into training and test sets.
4. Trained and compared two classification models: Random Forest / Decision Tree, and K-Nearest Neighbors (KNN).
5. Selected the best-performing model and used it to classify a new, unseen batch of customers.

Results

The Random Forest model was selected as the final model due to its higher accuracy and stronger resistance to overfitting on this dataset, and was used to classify new customers into Good, Standard, or Poor credit score categories.

Tools