Overview
This project builds a machine learning system that predicts whether an insurance claim will be made for an accident vehicle and how severe that claim is likely to be. The goal is to help insurers assess risk more accurately, speed up claims handling, and reduce manual work in the claims process.
Approach
- Collected and cleaned accident and policy data, handling missing values and outliers.
- Engineered features from accident type, vehicle information, location, and conditions.
- Trained and compared multiple supervised models including Decision Tree, Random Forest, and Support Vector Machine.
- Evaluated models using accuracy, precision, recall, and F1-score and selected the best performing model.
- Integrated the final model into a simple interface so that a new accident record can be scored in real time.
Impact
Working on this project gave experience in building an end‑to‑end ML pipeline: from data cleaning and feature engineering to model selection and deployment. It also showed how machine learning can directly support business goals in the insurance domain by making claims decisions faster, more consistent, and more data‑driven.
Sample outputs
Demo video & visuals
Demo of the insurance claim app from image upload to model prediction
Sample prediction where the model detects bumper damage on a vehicle image and labels the severity