About this project

Online Payment Fraud Detection System is a Python machine learning project for classifying online payment transactions as legitimate or fraudulent. The repository provides a Jupyter notebook and a Python script, along with a dataset folder, generated images, requirements.txt, README, and an MIT license. The documented workflow covers loading a transaction dataset, preprocessing, handling missing values, encoding categorical data, feature scaling, train-test split, model training, prediction, and model evaluation. The dataset fields described include step, type, amount, nameOrig, oldbalanceOrg, newbalanceOrig, nameDest, oldbalanceDest, newbalanceDest, isFraud, and isFlaggedFraud. The target variable is isFraud, where 0 represents a legitimate transaction and 1 represents a fraudulent transaction. Three classification algorithms are used: Logistic Regression, Decision Tree Classifier, and Random Forest Classifier. Evaluation is performed with accuracy score, confusion matrix, and classification report. The README includes a model comparison table and reports accuracy values for the three models, stating that Random Forest achieved 96.67% accuracy in the project's results. Technology requirements include Python 3.x, pandas, NumPy, Matplotlib, Seaborn, and scikit-learn; Google Colab is also mentioned as an environment. To run the project, clone the repository, install dependencies with pip install -r requirements.txt, then execute the Python script or open the notebook and run all cells after uploading the CSV dataset. The README lists future improvement ideas such as deep learning based fraud detection, real-time fraud monitoring, a Flask or Streamlit web application, API integration, and explainable AI. This is a compact educational example of a supervised classification pipeline for fraud detection.