Insurance Premium Prediction

📊 Machine Learning Regression Project – Kaggle Submission

This project builds a regression model to predict insurance premium amounts using a real-world dataset of 2 million entries. The goal was to optimize model performance while maintaining interpretability and memory efficiency under resource constraints.

🧠 Project Highlights

• Trained a Random Forest model on 2M records with custom preprocessing and feature engineering
• Applied tailored imputation strategies for missing values and encoded high-cardinality categorical features
• Achieved ~1.15 RMSE on Kaggle's private leaderboard
• Visualized distribution patterns and evaluated prediction errors to understand model limitations

📈 Tools & Libraries

• Python, Pandas, NumPy, Scikit-learn, Matplotlib, Seaborn
• Jupyter Notebook (exploratory analysis and training)

📊 Key Metrics

• 📉 RMSE: ~1.15 (Kaggle)
• 📉 MAE: ~637
• 📈 Records Used: 2,000,000+

📂 Dataset

Due to file size constraints, the dataset is not included in this repository.

Please download it directly from the Kaggle competition page:
🔗 Playground Series - Season 4, Episode 12

Place the downloaded files (e.g., train.csv, test.csv) in the same directory as the notebook before running.

🚀 How to Run

1. Clone the repo
2. Download the dataset and place it in the same directory as the notebook
3. Make sure you have Python 3 installed and the following libraries:

• pandas
• numpy
• scikit-learn
• matplotlib
• seaborn

4. Launch the notebook: