Leveraging Product Characteristics for Online Collusive Detection in Big Data Transactions

Abstract

Online fraud transaction has been a big concern for e-business platform. As the development of big data technology, e-commerce users always evaluate the sellers according to the reputation scores supplied by the platform. The reason why the sellers prefer chasing high reputation scores is that high reputations always bring high profit to sellers. By collusion, fraudsters can acquire high reputation scores and it will attract more potential buyers. It has been a crucial task for the e-commerce website to recognizing the fake reputation information. E-commerce platforms try to solve this continued and growing problem by adopting data mining techniques. With the high development of the Internet of Things, big data plays a crucial role in economic society. Big data brings economic growth in different domains. It supplies support to the management and decision-making ability in e-business through analyzing operational data. In online commerce, the big data technology also helps in providing users with a fair and healthy reputation system, which improves the shopping experience. This paper aims to put forward a conceptual framework to extract the characteristics of fraud transaction, including individual- and transaction-related indicators. It also contains two product features: product type and product nature. The two features obviously enhance the accuracy of fraud detection. A real-world dataset is used to verify the effectiveness of the indicators in the detection model, which is put forward to recognize the fraud transactions from the legitimate ones.