Design And Realization Of Commodity Fraud Monitoring System Based On Emotion Calculation

With the rapid development of E-Commerce industry, Fraud Detection is also became a big problem for E-Commerce. Because of the information asymmetry, some of the online sellers will make spam product comment to raise their own reputation score or decrease others to cheat customer. So how to help customer identify these fraud seller become more and more important now.To solve this problem, this paper proposes a fraud detection system based on affective computing, which converts the text customer comments or product reviews to objective indicator to evaluate the credit of seller compared with social network comments and professional product comments.First of all, we collect product comment from target and benchmark by crawler or API; Then we mine product features from comments by association rules algorithm and product dictionary based on domain ontology to build the indicator system; After that, find out the feature-opinion pair, degree adverb and negatives by dependence relation, HowNet Emotional Corpus and word similarity calculation based on the indicator system. At last, reduce dimension based on semantic and identify fraud by outlier detection algorithm.The proposed fraud detection system based on affective computing start from the product URL and collect the product reviews data by pre-defined configuration. It then extract product features and corresponding opinions, and find out which one may be fraud to help customer make better shopping decision, which has high practical application value.