Research On Multi Attribute E-commerce Information Collection And Recommendation System

With the e-commerce platform developing rapid,more and more people choose to do shopping online.Consumers can buy goods that meet their needs without going out of their homes.In order to help consumers screen goods,the platform mainly through the consumer's history browsing record or a single attribute of the goods to recommend.But when consumers shopping online,they are always with pertinence and favoritism,and pay more and more attention to the evaluation of goods,and consider the attributes of different commodities to decide whether to buy goods.If consumers spend too much time and energy still choosing not to meet the needs of the goods,it will greatly reduce the shopping experience of the e-commerce platform.Therefore,in order to enable consumers to find products that meet their individual needs faster,better and more intuitively,based on commodity evaluation and multiple attribute characteristics of commodities,through the emotional analysis of the evaluation content and attribute selection,establishing a multi attribute personalized recommendation method,help consumers quickly find satisfactory goods.The main research contents include:First,the basic data of the evaluation and properties of the goods are obtained by writing a crawler based on Python.After preprocessing the evaluation data,the attributes of the goods and the evaluation words are extracted.By analyzing a large number of evaluation contents,we can get the grammar mode of Chinese evaluation,and extend the syntactic pattern for phrase Combination Based on the existing Chinese common patterns.The syntax tree is constructed combined with the statistical grammar mode,and then the commodity attributes and evaluation words are extracted through matching the syntax tree,and the extracted commodity attribute names are normalized.Secondly,based on the product evaluation data,we build the corpus of our own domain,distinguish the positive and negative faces of the evaluation,use the SnowNLP model to train the corpus,get the new model,then use the model to process the evaluation content and get the emotional value.Among them,each attribute corresponds to the different emotional values.Consumers give different priority to the evaluation attributes and use the weighted summation of emotions to calculate the final emotional value of the commodity evaluation.Then,in order to enable consumers to consider many attributes of the goods(including evaluation emotion)quickly and intuitively,a multi attribute decision making method is used.By assigning different priority of commodity attributes,consumers use the corresponding attribute values to calculate the comprehensive score of goods using weighted summation,which is used as a commodity recommendation index.Finally,in order to complete the verification,this paper designs and implements a recommendation system prototype based on this method,and recommends the products in the fruit field of Jingdong platform.The results show that the method can for consumers of different commodities,different attribute priority personalized settings,evaluation and attributes based on the recommendation made different results,help to meet consumer expectations of rapid positioning of goods,has a certain practical value.