Fraud Detection on Crowdfunding Platforms Using Multiple Feature Selection Methods

Abstract

Crowdfunding platforms have emerged as an effective medium for collecting financial support from individuals, businesses, and organizations for various innovative ideas, social causes, medical needs, and startup projects. However, the increasing popularity of online crowdfunding has also resulted in a rise in fraudulent activities, where fake campaigns are created to deceive donors and misuse collected funds. Detecting such fraudulent campaigns manually is a challenging and time-consuming process due to the large volume of data generated on these platforms. This project proposes a machine learning-based fraud detection system for crowdfunding platforms using multiple feature selection methods to improve the accuracy and efficiency of fraud identification. The proposed system analyzes various campaign-related attributes, including campaign description, funding goals, creator details, social engagement, donation patterns, and historical activities, to identify hidden fraud patterns. Multiple feature selection techniques are applied to extract the most significant features from the dataset, reduce irrelevant information, and enhance the performance of machine learning models. Classification algorithms such as Random Forest, Support Vector Machine, Logistic Regression, and XGBoost are utilized to categorize crowdfunding campaigns as genuine or fraudulent. By combining effective feature selection strategies with advanced machine learning techniques, the proposed approach aims to provide a reliable, automated, and accurate fraud detection mechanism that helps crowdfunding platforms minimize financial risks, protect donors, and increase trust in online fundraising environments.