The Algorithm And Implementation Of Cross-Platform Personalized Recommendation Based On Association Rules

Accurate andefficient recommendation system could be used for mining user’spotential selection inclination, whichwill provide personalized service for many users. With the maturity of the mobile internet and integration of the corporate business, a personalized recommendation system which is cross-platform will greatly increase user viscosity as well as improve the user experience. Recently the research of personalized recommendation is mainly centered on improving the efficiency or accuracy of the recommendation algorithm, which pays relatively little consideration to the cross-platform, cross-domain personalized recommendation needs. Therefore considering the cross-platform user experience needs, this paper propose a personalized recommendation system design based on association rules including following aspects.Firstly, this paper first proposes a cloud storage protocol using cookie single-sign-on technology to achieve cross-platform, cross-domain data storage. According to the design of storage protocol, the personalized recommendation system willrequire high-speed return of recommendation results. However, the traditional personalized recommendation algorithm based on content must maintain all the user profiles and search through all the products or the page to get the best recommendation each time, which is not applicable for personalized recommendation system proposed in this paper.Secondly, this paper proposes a recommendation algorithm of association rules based personalized recommendation system, to satisfy the real-time demand of returning results. Firstly, this paper uses the frequent pattern recognition algorithm for non-real time update to establish frequent pattern database. Secondly, in order to reduce the difficulty of real-time personalized recommendation, the feature mapping and cluster analysis are applied to sequence in the frequent pattern database. Finally, according to the clustering results, a specific evaluation standard for each class is established, and thus completing the design of the whole cloud personalized recommendation system.Finally, based on these methods, this paper applies algorithm to the real internet system to test accuracy and efficiency by mining real user sequences.Considering the cross-platform personalized recommendation needs, this article puts forward a new design of the cloud storage, and presents thematchingdata processing algorithm. This algorithm not only consider the actual system demand but also takes cluster characteristics of Internet users into account, which improve the overall accuracy. To this extent, this paper is meaningful to improving the internet user experience and integrating internet service.