关键词: 商品评价综合评分, 矛盾性评价, LSTM-AE, 支持向量机

Abstract:

 Commodity evaluation from e-business platform and mobile application store is an important reference for consumers However, under current methods, the existence of contradictory evaluation will lead to distortion of commodity scoring and interfere with consumers′ shopping decisions In order to solve this problem, a comprehensive scoring evaluation method is proposed for both comprehensive text reviews and numerical scoring The LSTM-AE neural network is introduced to extract the implicit features of text reviews, based on which, the text score is obtained by Support Vector Machine (SVM)And then the score is weighted with the numerical scoring to evaluate the comprehensive score Using the model that combines LSTM-AE neural network and SVM to classify the commodity evaluation data from JDcom, the accuracy rate can reach 8870% Through using the mentioned method to make comprehensive score of evaluation from multiple sources, it is shown that the method can neutralize the contradiction of the differences in the contradictory evaluation, and meanwhile it doesn′t affect the score of non-contradictory evaluation This method has certain feasibility

Key words: comprehensive score of evaluation, contradictory evaluation, LSTM-AE, SVM