Fraud detection model for illegitimate transactions

Abstract

Due to advancements in network technologies, digital security is becoming a top priority worldwide.  This project aims to study how machine learning techniques can be used to learn patterns in fraudulent and legitimate transactions in order to detect fraudulent transactions using Python programming language on Jupyter notebook as the integrated development environment (IDE). Scikit-learn was used to process the algorithm, and Streamlit and Heroku platforms were used for deployment of the algorithms.  This was incorporated into a web application that allows the user to upload data that is analyzed by the system to detect fraud. The Classification report and Confusion matrix are used to evaluate each model’s accuracy. The random forest model gave an accuracy of 99.95 %. At the end of this study, a web-based application was developed to allow users upload data and also to detect fraudulent online based transaction.