| Nowadays the e-commerce trading platform is developing in a rapid way, while the logistics system plays the supporting role. However,online comment function set up to make up for the lack of online shopping perception, cannot effectively prevent the emergency of spam reviews. Spam reviews have a direct effect on unnecessary costs like delivering empty package, misleading purchase, human resource and material consumption. At present, the logistics industry and e-commerce platform share weal and woe. Thus how to detect the spam review on trading platform in front of the logistics industry is of great significance to reduce logistics costs and standardize the entire e-commerce market.This research applies the Web crawler to collect and pre-process real data from the trading platform in order to ensure the applicability of the result. E-commerce platform tends to show personalized reviewing and common purchasing behavior, while the two behaviors are actually coherent. Therefore, after studying related references, this research firstly divides four typed of products, and shows the consumer behavior(including both reviewing and purchasing behavior for one consumer)under continuous date. Combined with analysis of time difference and different dimensions, the research identified the suspicious spam review groups. Secondly, this research carries on text analysis to the review content, including constructing the sentiment lexicon, applying word segmentation, applying the Apriori algorithm to extract the text theme,constructing the text feature vector and calculating the text similarity.Lastly, this paper combines data analysis of consuming behavior with text analysis of reviewing behavior, and proposes a detection method for spam review using K-means algorithm. The research can be applied to the automatic collection of online review and detection of spam reviews, and shows both efficiency and effect. |
PDF Full Text Request