Research And Implementation Of Personalized Recommendation System Based On Association Rules And Clusteirng Analysis

With the rapid development of e-commerce, especially with the constantlyexpanding scale of mobile E-Commerce, consumers are submerged in an ocean witha huge number and variety of merchandise. During the period of shopping,users willvisit a large number of useless commodities or commodities which they are notinterested in. Before users finally find the goods they need, they will spend a lot oftime and effort to filter out the goods which they are not interested in or do notmeet their demands. This will cause a problem called information overload,and willaffect consumers’ shopping experience, and ultimately lead to the loss of consumers.To solve these problems, personalized recommendation system came into being.Personalized recommendation is to recommend commodities of interest to the users.Personalized recommendation systems need to collect users’ interests, and customizepersonalized content for users based on these data, and finally generate therecommendation result, this is a key element when researching personalizedrecommendation technology.Personalized recommendation technology has become the core of many matureE-Commerce systems, there are many excellent and mature recommendationtechnologies having been widely used in traditional E-Commerce systems, such ascollaborative filtering technology, association rules recommendation technology,clustering analysis recommendation technology. There are still some differences between mobile E-Commerce system and traditional E-Commerce system, butresearch on personalized recommendation for mobile E-Commerce system are few.Mobility and personalization are two main characteristics of MobileE-Commerce that are not available in traditional E-Commerce. Meanwhile, with thecontinuous development of electronic commerce, there are urgent need forrecommendation system to provide users with dynamic recommendation servicewhich can recommend accurate commodity to provide users with “compare andrefer” services to provide users with real-time optimization of distribution servicesand other services, all of these requirements need to be solved. And the advent ofmobile E-Commerce provides an unprecedented opportunity to achieve thisrequirement, which makes mobile E-Commerce become a new direction ofE-Commerce development. But this also brings a series of problems to be solved forbusiness intelligence technology of mobile E-Commerce,especially personalizedrecommendation technology.This topic made a deep research on personalized recommendation technologyand related theories of mobile E-Commerce at the same time analyzed users’consumption patterns under the mobile environment; on the basis of previousstudies, focused on clustering analysis algorithms and association rule algorithms,after analyzing the advantages and disadvantages of the two algorithms,proposed animproved algorithm combining the two algorithms which could improve the qualityof personalized recommendation and could improve the accuracy ofrecommendation; based on the theories and methods in this paper, designed andimplemented a personalized recommendation system in mobile electronic businessarea, this recommendation system used the improved algorithm, and wereapplicated into a real commercial project of mobile E-Commerce. This paper designed and implemented a personalized recommendation systemfor mobile E-Commerce, this system used combining algorithm which was based onassociation rules and clustering analysis algorithm to recommend products for user.This system consisted of two modules: offline mining module and onlinerecommendation module. Among them, the offline mining module used clusteranalysis algorithm to parse the log file, and classified the result, and then usedassociation rules algorithm to generate association rules based on classified data,finally, system used the user’s browsing behavior data to match the set of rules whichthe users belonging to, generated the recommendation result; the onlinerecommendation module processed users’ requests and showed therecommendation result to users.