Research And Implementation Of Mining Individual Cognitive Value Based On The Network Comments

With the rapid development of computer science and the popularity of the social networking service, e-commerce has become an important part in people's daily life. E-commerce network has evolved into a special social network with social characteristics. When shopping online, netizen tends to make various comments as well as social interaction with the merchants and other peers. These comments influence potential consumers a lot in their decisions. Therefore, it is of great significance to analyze the online behaviors of the individual, and identify the user emotions, influence, and cognitive value from massive comments information. The research on individual cognitive value in this thesis is a part of the 973 project “Individual Behavior Analysis of Social Network”. The key technology can also be extended to individual cognitive value analysis of social network. The main contributions including technology research and project implementation are listed as follows:(1)Technology research: We propose a text mining method of individual cognitive value influence factors based on online comment data. Firstly, a consumer cognitive value model is established. Then comments from consumers are preprocessed. The similarities among comments are calculated by angle cosine. After that, we apply K-means to cluster all the comments. The attributes word extraction method based on differences is proposed as well. It is able to extract representative attribute keywords from comment collection by comparing comments between popular and unsalable products. On this basis, we cluster all the word attributes using frequent items mining and then obtain an ordered set which contains high positive impact factor attribute words.(2)Project implementation: We design and implement a prototype system of individual consumer cognitive value mining. The system consists of three modules including data collection module, data storage and pretreatment module, and data analysis module. We grab the seller transaction data of dendrobium candidum(a kind of health care products) from Taobao(The first three quarters of 2014,251129 comments). After comments clustering, attribute words extraction and frequent items mining, 67 ordered sets of cognitive value attribute words including efficacy,price and service have been obtained. The artificial cross sampling test shows that the above experimental accuracy is 86% and the recall rate is 84%, which verifies the validity of our method.