Personalized Recommendation Based On The Mobile User Of Behavior Transfer Pattern

The rapid development of mobile Internet technology and the gradual penetration and popularization in all walks of life of computer intelligence,greatly changing the user's behavior pattern.Today,the research of mobile user's behavior pattern and personalized recommendation has become the key technology of the e-commerce companies to compete market share.In view of the context awareness feature of the mobile user's behavior sequence,a research method of mobile user behavior transfer pattern and personalized recommendation is proposed in this paper,which mainly completes the following research work.Firstly,a data preprocessing method is proposed,and this method contains two aspects of content:(1)In view of the user's behavior in mobile environment has the sensitivity of time and position,the data preprocessing algorithm TP-constrain is proposed to eliminate noise.For the transaction database generated after TP-constrain processing,this thesis put forward a method for converting data,that is to convert transaction database into decision table.The TP-constrain method achieves the purpose of getting more precise samples by eliminating noise,and the experimental results show that this method is effective and more accurate in the mining of user sequence behavior pattern.Secondly,a mobile personalized recommendation algorithm(MPRC)based on context awareness is proposed.This recommendation algorithm is mainly divided into two steps,the first is to excavate the mobile user behavior transfer pattern based on context awareness,the second is to rank the pattern,and the result of the ranking is considered the influence factor of the contextual similarity.The main work is as follows:(1)This algorithm uses Apriori to filter the data of user's historical behavior in order to obtain a frequent pattern with length 2,and then converts these data into the decision tables.Finally,using the method of attribute reduction of rough set to mine transfer pattern of context awareness from decision tables.(2)Pattern sort based on the contextual similarity is proposed.The pattern sort method not only depends on the support and confidence,but also need to consider the contextual similarity.Based on this,a pattern sort method based on the contextual similarity is proposed.By comparing the experimental results of MPRC and multidimensional approach for context-aware recommendation,it is proved that the MPRCalgorithm has a higher accuracy in the mobile environment.