Research On Personalized Review Ranking Method Based On Consumers' Multidimensional Preferences

With the booming development of e-commerce and user-generated content,the massive online reviews accumulated on websites have exceeded the information processing capacity of consumers.Due to the severe overload of information,consumers can only choose to browse part of reviews in the order of display,which is always unscientific.Hence it is particularly necessary to explore new ways to rank product reviews.Existing review ranking methods are basically oriented to consumer group.However,with the popularization of big data technology and customized services,it is obviously more in line with the market development trend to provide personalized review rankings for consumer individuals.A few existing personalized methods only consider single-dimension consumer preference.Thus,how to mine consumers' multidimensional preferences to generate personalized review ranking is the problem to be solved in the thesis.Firstly,it is the basis of personalized review ranking to comprehensively depict consumers' preferences from multiple dimensions.In the thesis,the concepts of product feature preference,review sentiment preference and browse quantity preference are put forward by analyzing consumers' behavioral habits and interested information in the process of browsing reviews.Also,the mining and quantifying methods of each preference are analyzed in detail.Secondly,multidimensional preferences are added to the review ranking model,and the role of each consumer preference in the model is determined according to its characteristic.Then,we propose the consumer preference satisfaction measurement method of a single review,a review collection and a review ranking in turn,so as to calculate the expected satisfaction of all possible review ranking lists.In this way,we transform the ranking issue into an optimization problem,whose goal is to maximize the expected satisfaction,thus obtaining the corresponding optimal review ranking.An example is given to demonstrate the process of using this model to sort reviews,and the necessity of considering multidimensional consumer preferences is verified by comparing with methods only considering single-dimension consumer preference.Finally,we analyze the complexity of the optimization problem as a NP-hard problem,and propose a heuristic approximate algorithm for solution.Intensive experiments areconducted on real data from a famous hotel website in China,and the optimal parameter of the algorithm is determined.The results reveal that the proposed approach has the best effectiveness compared with other relevant methods and high sensitivity.The study has certain practical significance.Consumers can perceive the real quality of products and make purchase decisions scientifically by browsing review ranking information accorded with their own preferences.According to consumers' mainstream preferences and popular reviews,business can optimize the design of products and make relevant marketing strategies,in order to stimulate product sales and improve market competitiveness.