The Research Of Method To Obtain The Individualized Demand Based On The Online Product Evaluation

Online individual product customization has been favored by more and more businesses and consumers on account of its advantages such as fast response, convenience of purchase and sale, relatively low price and so forth. It is a key link to gain the real individualized demand accurately and comprehensively. The traditional ways, such as questionnaire and interviews, cannot obtain some information accurately and fully as a result of human cost, location and other restrictions. With the development of e-commerce technology, online shopping gradually becomes the new choice of people. Then the enormous and multiple after-sales evaluation, which are very important basis for a lot of consumers to choose the business and the brand and for the manufacturing enterprises to improve and develop new products, are of precious value for commerce and research. Therefore, it is meaningful and significant in terms of theory and practice to gain individualized demand from the online product evaluation.Currently, the mining of online product evaluation focuses on identifying topics, extracting features of products, mining viewpoint words, determining attitudes polar, and so on. However, there still exist some deficiencies in the process of mining enormous evaluation data:(1) Due to the openness of the Internet and the freedom of consumer expression, online product evaluation data contains a lot of information useless for acquiring the consumers’individualized demand. If no preprocessing were conducted to the valueless evaluations, the quality of acquiring the real individualized demand would be seriously affected.(2) Previous studies pay too much attention to factual evaluation or evaluation about opinions, ignoring information redundancy, which significantly increases the database information retrieval workload. Based on the theoretical knowledge and methods of product design, consumer individualized demand acquisition, the mining of online product evaluation, this paper puts forward a binary classification method of online evaluation information, a processing method of redundancy on individualized demand acquirement based on novelty, a method of gain consumers’individualized demand based on polarity words’emotional tendency. And through the example validation, these methods can gain individualized demand more accurately and comprehensively.1. Recognizing trash evaluation on online product evaluation. This paper defines and classifies trash evaluation, and regards the recognition as the sublevel problem of the mining of online product evaluation. It means to classify the evaluation information as useless and useful, and then use SVM to realize the recognition of trash evaluation, so as to preprocess the data before obtaining the individualized demand more accurately.2. The research of method for dealing with the redundant information to obtain individualized demand based on novelty. Through analyzing the relationship between keywords and polarity words, this paper designs the algorithm process to extract the subject headings and the polarity word. This paper also puts forward and designs the mining model and the algorithm for the redundant information to reduce the workload of information retrieval.3. The research of method to gain consumers’individualized demand based on polarity words’emotional tendency. A polarity word dictionary of acquisition is compiled with the help of HowNet. Therefore, a design is conducted to the algorithm procedure of emotional tendency of polarity words. Those derogatory words are selected according to the discrimination results of polarity word’s emotional tendency; Algorithm procedure of requirement elicitation is designed to acquire individualized demand more accurately and comprehensively.4. Finally, this paper does the instance validation, sets up the experiment environment, and analyzes the result. The proposed method is verified by the best-selling smart-phone evaluation records that are obtained from Taobao.com through crawlers written in Java. Through the feature vectors choosing in this paper, the accuracy of the algorithm arrives89.2%. In the experiment of processing the redundant information based on novelty, the best performance of the algorithm reached93.2%after a set of different novelty threshold. The experimental result of the polarity words’emotional tendency indicate that the accuracy of the method used in this paper to recognize the commendatory polarity words reached90.3%, the accuracy of the method in recognizing the derogatory polarity words reached81.6%.