Transplant Track: Intelligent Detection of Online Payment Frauds Using Machine Learning Techniques

The growth of online bank and digital payment systems has been a major contributor to credit card fraud that causes financial losses in a big proportion to both the financial institutions and the end users. The use of transaction datasets can be seen as highly imbalanced, which makes it difficult to detect fraudulent transactions because fraudsters keep using new strategies. This study offers a credit card fraud detector model with machine learning and deep learning to develop an effective credit card fraud detector that avoids false positives and identifies suspicious transactions. The training and evaluation are done using a publicly available credit card transaction dataset. Preprocessing data techniques like normalization, feature selection, and sampling techniques are used to solve the imbalance of the classes. Various classification algorithms such as the Decision Tree, the Random Forest and Support Vector machine are executed and compared based on evaluation measures such as accuracy, precision, recall, and F1-score. The most efficient model is combined to create a prototype web-based application which is created with the help of HTML, CSS, Bootstrap, and Flask to make a real-time prediction of the fraud. The suggested system enhances the security of the transactions and this offers a scalability base in the future in case of increase like deploying the cloud service and automatic alert system.