🛡️ Insurance Fraud Detection using Machine Learning

This project is focused on building a machine learning model to detect fraudulent financial transactions based on patterns in the data.

📌 Problem Statement

To detect and classify insurance claims as fraudulent or genuine using supervised learning techniques. Insurance frauds are often hidden and lead to huge financial losses in the insurance sector.
The goal is to train a machine learning model to recognize patterns that typically indicate fraud.

🧠 Technologies & Libraries Used

• Python
• Pandas
• NumPy
• Scikit-learn
• Jupyter Notebook
• Matplotlib / Seaborn (for data visualization)

📊 Project Workflow

1. Data Loading & Preprocessing

   • Handled missing/null values
   • Feature selection and encoding
   • Scaling and normalization

2. Exploratory Data Analysis

   • Distribution plots
   • Class imbalance analysis
   • Correlation matrix

3. Model Building

   • Used Logistic Regression for classification
   • Split data into training and test sets
   • Fit the model and evaluated results

4. Evaluation Metrics

   • Accuracy Score
   • Confusion Matrix
   • Precision & Recall

🚀 How to Run

1. Clone this repository:

   git clone https://github.com/Adilkhan6465/fraud_detection_project.git
   

📈 Results

• Accuracy Achieved: 93.12%
• The model shows strong performance in identifying fraudulent transactions.
• Further tuning and use of ensemble models may improve performance slightly.

folder structure

fraud-detection-project/

├── data/ --> all datasets │ ├── insurance_data.csv │ ├── fraud data FY 2023-24.csv │ └── feature_engineered_fraud_data.csv │ ├── models/ --> save trained model │ └── fraud_detection_model.pkl │ ├── scripts/ --> (training, preprocessing etc.) │ ├── model_training.py │ ├── data_preprocessing.py │ ├── feature_engineering.py │ └── app.py │ ├── test_api.py --> API test ├── requirements.txt --> Python libraries list ├── README.md -->

🔮 Future Improvements

• Implement advanced models like Random Forest, XGBoost, or SVM
• Handle class imbalance using SMOTE or undersampling
• Create a web interface using Flask/Streamlit for real-time predictions
• Add model explainability (SHAP/LIME)

👨‍💻 Author

Adil Khan
GitHub: github.com/Adilkhan6465

---