Research On User Behavior Based On Commodity Review Information

In recent years,shopping on e-commerce platforms has increasingly become a major channel for people to purchase goods.With the increasing use of e-commerce platforms by users,massive amounts of data on user purchase behaviors are left on e-commerce platforms.Mining and analyzing these behavioral data is very important for both users and businesses.This article analyzes the massive behavioral data of the e-commerce platform to obtain information about the user's purchase of goods,the user's community preference,and the user's purchase tendency.This information can help merchants carry out product promotion and warehouse stocking,and can also provide users with information Need to purchase goods to provide convenience.Based on this,this article mainly conducts research from three aspects,as follows:1.In order to solve the problem of data sparsity when traditional topic model is applied to short text topic mining due to the lack of sufficient context information,this thesis proposes a short text topic model algorithm based on semantic enhancement.In this algorithm,DMM is combined with a word embedding model.The vector representation of words is obtained by training global and local word embedding,respectively.Subsequently,the semantic correlation between the global and local word embedding vectors is calculated by fusing the global and local word embedding vectors,and the semantic-related word sets of subject words are constructed.And calculate the weight of related words of the subject words;Then,the semantic enhancement of words is calculated by the weight of theme-related words,and finally the topic model of words is mined.The experimental results show that the proposed topic model is more accurate in the representation of topic consistency,and improves the accuracy of the model in the classification of short texts.2.Aiming at the inaccuracy of community division caused by the selection of truncation distance and only considering the topology of community network in the application of density peak clustering algorithm in community division,a community detection algorithm of density peak is proposed,which integrates the node attributes of users and the topology of the network in which they are located.Firstly,the user network topology is used to calculate the degree between the user node and its direct and indirect neighbors,and the degree is used to represent the local density between nodes.Then,the user's preference for the comment information topic is taken as the attribute of the user node,and the similarity between users is calculated by combining with the network topology structure and the relative distance between users is expressed.Finally,the key nodes are selected as the central nodes of the community and the community division is completed.Experiments show that the proposed algorithm is superior to the baseline model algorithm in both ACC and NMI indexes,and improves the accuracy of the community detection algorithm on the e-commerce network platform,and realizes the efficient community division.3.In view of the problem that most of the current user purchasing behavior prediction only uses the user's browsing and viewing operation of goods,the behavior of collecting goods and adding goods to the shopping cart,but rarely analyzes the user's comment information on the e-commerce platform.This paper proposes a method to predict the purchasing behavior of users based on the emotional polarity of user reviews and user community influence.Firstly,the dictionary-based sentiment analysis method was used to obtain the user's emotional polarity of purchasing products,and then the user's influence in the community was obtained.The fuzzy-based technology is used to integrate the user's purchasing emotional polarity and community influence,so as to calculate the user's purchasing emotional polarity finally.Finally,the product purchase behavior of users is predicted according to the calculation results.Experimental results show that the method proposed in this chapter can obtain the emotional polarity conveyed by users more objectively and accurately than the traditional method,and has better prediction accuracy than the baseline model in the aspect of user purchasing behavior prediction.