Design And Implementation Of C-eommerce Recommendation System Based On Data Mining

Consumers’ choose breadth and depth is increasing with the development ofe-commerce. Consumers are motivated to but by E-commerce, meanwhile, they have toface to mass goods and fell difficult to choose the products they want, even make theminto a data rich and poor knowledge of the situation.Recommended system aims simulate real-life salesman recommended product toconsumers, and help consumers find their own satisfaction goods. Existing e-commercerecommendation algorithm is based on user browsing behavior and uses the users’similarity behavior as recommended; however, this process ignores the user’s emotionalattitude of commodities. In fact, the positive sentiment orientation will be increasedcustomers’ satisfaction, and vice versa.In order to improve user satisfaction of the recommendation results, this thesisproposed an algorithm which fuses the sentiment orientation mining based on studyingthe existing recommendation algorithm. In this method, the users have the samebehavior if they are in the same categories. Base on this assumption, a candidate list ofrecommendation results is produced. In recommendation processing, the candidate listof recommendation results is filtered by the consumers’ sentiment orientation, removethe poor evaluation products and reserve the fine evaluation products. This thesiscollected the users’ reviews of products, and analyze theirs sentiment orientation basedon sentiment words. In order to generate the sentiment word, this thesis consider thatthe text containing more positive emotion words is positive, meanwhile the wordscontained in the more positive text is positive, and vice versa. And, the point mutualinformation method is used to extend sentiment words. The experiments prove theproposed recommendation algorithm helps improve user satisfaction.