A Research of Taobao Cheater Detection

Abstract : This paper focuses on Taobao cheater detection. At present the phenomenon of fake trading is widespread in Taobao, which makes it difficult for consumers to distinguish between true and fake product reviews. To solve this problem, we collect a total number of 50,285 historical review data from 100 cheaters and 100 real buyers to create a dataset. By using these data, we extract 8 features from three dimensions that are reviewer, commodity, and review. Then we use the SVM algorithm to construct the classification model and choose the RFB kernel function, which has a better performance to identify the cheater. The precision of the final classification model we built to identify the cheater reaches up to 89%. The experimental result shows that extracting features from the historical review data can recognize the cheaters effectively. It can be applied to the recognition of the cheaters in Taobao.