Research On B2c Customer Consumption Preference Model Based On Review Data

The rapid development of the Internet boosts the quickly increasing of shopping online.Recent year,with the development of B2C shopping mode like Jingdong and Tmall,the rapid growth of the volume of business and information brings challenge to the development of enterprise.How to purify useful information from the increasing unstructured data?How to excavate the true demand of customer from mass consumptive data to offer accurate personalized service and improve the shopping experience furthest?These questions have become the hot topic and difficulty at present.So,the mode of data driver can be used to excavate the consumption preference of customer,and it is the important safeguard guarantee for B2C shopping website to do precision marketing.This paper based on the theory of online review,consumer behavior and the customer consumption preference of B2C shopping website.Tmall clothing consumption has been chosen as research object,influence factors of customer consumption preference have been analyzed according to consumer,platform and merchant,and the target products have been classified and screened to ascertain 7 kinds of clothing products,the crawler software was used to capture the online review from September to November in 2016.By the way of data collating,key words extracting and statistic analyzing,34 high frequency of consumer review information have been excavated and 12 characteristic factors of variables have been ascertained.5 grade rating scale of Likert scale was used to translate the comment information into structured data.The Clementine 12.0 software was used to import 12 characteristic factors of variables of products to build Bayesian network of factors.The conditional probability distribution of each node under the parent nodes and the importance of each characteristic factor have been calculated.Logistic regression model has been build to compare the accuracy with Bayesian network model to do accurate assessment.The research shows that the comfort degree,fabric work,color,quality,suitable degree and price of 7 selected products are high frequency focused words of consumer;there are strong correlation of nodes in Bayesian network model;the conditional probability distribution of nodes are similar,the probability of reviews like excellent,good and general gave by consumer is higher;the importance of characteristic factor of man's clothing and woman's clothing are different.Woman's clothing focus on logistic,coincidence degree,handle,authentic,suitable degree,man's clothing focus on fabric work,color,logistic,handle,appearance degree and comfort degree.Consumption preference can be analyzed and precision marketing strategy can be made by B2C shopping website according to high frequency focused words,correlation of factors in Bayesian network,conditional probability distribution and importance of factors.