온라인 C2C 중고거래 시장에서의 사기 탐지 연구

Abstract

With the prevalent use of information and communication technologies, the volume of online C2C transactions has largely increased. Among online C2C markets, the demand for C2C second-hand market is particularly growing as the increasing number of consumers want reasonable consumption by using this market. However, as the volume of transactions increases, the number of frauds, which intend to earn unfair gains by sending products different from specified ones or not sending them to buyers, is also increasing. Therefore, the solutions for finding and preventing fraud are highly needed. For this reason, this study explores the model that can find out the frauds in the online C2C second-hand market by examining the postings for transactions. The model is built with the characteristics from postings such as the topic and the linguistic characteristics of the product description, and the characteristics of products, postings, sellers, and transactions. For extracting topics from postings, LDA is used. For the linguistic features of the product description, the number of specific parts of speech, the length of texts, the rate of white spaces, line breaks, and special characters in the writings are used. For product characteristics, the rate of zeros in the price is used. For the posting characteristics, the day and time of the posting and the number of images are used. For seller characteristics, the grade of the seller and the sharing of the email address are used. For transaction characteristics, whether to use a secure transaction is used. The constructed model is then trained by the two machine learning algorithms: XGBoost and Deep learning. Then their performance of detecting fraudulent posting is evaluated and compared. As a result, the model that uses XGBoost showed a higher performance thus selected to provide the detailed findings of exploring fraudulent postings. The final analysis results show that fraudulent postings have less information, which is also less specific, fewer nouns and images, a higher ratio of the number and white space, and the shorter length than genuine postings do. Also, unlike the fraudulent postings, the genuine postings are focused on the product information, product evaluation, and delivery information. This study shows that the various features can be extracted from postings written in C2C second-hand transactions and be used to construct an effective model for frauds. The proposed model can be also considered and applied for the other C2C platforms. Overall, the model proposed in this study can be expected to have positive effects on suppressing and preventing fraudulent behavior in online C2C markets.