This project focuses on building a machine learning model to detect fraudulent credit card transactions using historical transaction data from the USA. The dataset is highly imbalanced, and various techniques were employed to ensure accurate and efficient fraud detection.
The goal is to predict whether a credit card transaction is fraudulent based on transaction features, helping financial institutions reduce fraud and improve security measures.
- Code: Python scripts for data preprocessing, exploratory data analysis, and model building.
- Datasets: Processed datasets used for training and evaluation.
- Power BI Dashboard: Visualizations summarizing customer churn analysis.
The Power BI Dashboard file is too large to include directly in the repository. It is available in the Releases section of this repository.
- Navigate to the Releases section to download the Power BI file.
- Click here to go to Releases
- Data Preprocessing: Cleaned the dataset and performed feature engineering using Pandas and NumPy.
- Data Balancing: Applied SMOTE (Synthetic Minority Over-sampling Technique) to balance the imbalanced dataset and improve model performance.
- Modeling: Built and trained multiple classification models, including Support Vector Machine (SVM) , optimizing them with Cross- Validation.
- Model Evaluation: Achieved an 84% accuracy in predicting fraudulent transactions while minimizing false positives.
- Python (for model development and data manipulation)
- Scikit-learn (for machine learning al
- Scikit-learn (for machine learning algorithms)
- Pandas and NumPy (for data processing)
- Power BI (for visualizing fraud detection insights)
- SMOTE (for data balancing)
- Matplotlib (for data visualization)
- Successfully detected fraudulent transactions with 84% accuracy.
- Reduced false positives and improved fraud detection efficiency.
- Visualized fraud trends through a Power BI dashboard.