In today’s world where acts of kindness are seldom and rare, there are still many people who are able and willing to help their fellow human beings. One such way of doing that is donating to a crowdfunding campaign. People in need of financial assistance describe their stories on a crowdfunding platform and generous people donate to these campaigns. Even in such a noble cause, there are malicious actors who post fake campaigns and misuse the donations made to the campaign. In this study, we propose a fraud detection method to classify a campaign as genuine or fake. We have collected the details of non-fraudulent campaigns from ww.GoFundMe.com and we are collecting details of fraudulent campaigns from www.GoFraudMe.com. We propose a combination of machine learning classifier and a rule-based classifier to classify a campaign as genuine or fake. We have based our rule-based classifier on theories in deception which uses cognitive load, certainty, emotion, and distancing strategy depicted in a text. We then aggregate the results of these two classifiers to label a campaign as genuine or fake. Fraudulent campaigns add up to $30M and hence their detection has significant practical use.
Prateek, Pranay; Kim, Dan J.; and Ge, Ling, "Detection of fraudulent campaigns on donation-based crowdfunding platforms using a combination of machine" (2021). WISP 2021 Proceedings. 2.