基于频繁项集的多用户数据流混合推荐仿真

Multi-User Data Stream Hybrid Recommendation Simulation Based on Frequent Itemsets

针对当前数据流推荐存在安全性差和精确度低的问题,提出基于频繁项集的电商多用户数据流混合推荐方法。将蚁群算法与属性相关分析融合,将蚁群收敛所至路径判断为存在安全隐患的路径,对路径中存在安全隐患的数据进行判断,确定最终异常值并去除。基于数据流安全分析,引入增量挖掘算法,采用次频繁项索引表更新频繁树,并利用压缩FP-树与矩阵技术对新频繁树的频繁模式进行挖掘。根据数据流频繁模式挖掘,将不同数据流分类至相应主题组。基于频繁项集计算多用户访问相似程度,同时找到用户最近邻平时访问数据。结合主题抽取、最近邻访问及多用户共同兴趣相似度计算,将最符合用户的数据流推荐给用户,实现用户数据流混合推荐。实验结果表明,上述方法推荐过程安全性能好,且推荐结果准确度高,是一种切实可行的数据推荐方法。

At present,the data stream recommendation leads to low security and low precision.Therefore,this paper focuses on a method of hybrid recommendation for e-commerce multi-user data stream based on frequent item set.The ant colony algorithm was integrated with the attribute correlation analysis.And then,the path of ant colony convergence was judged as the path with security risk.The data with security risk in the path was judged,so that the final abnormal value was determined and removed.Based on the data flow security analysis,the incremental mining algorithm was introduced and the frequent tree was updated by the sub-frequent index table.Meanwhile,the compressed FP-tree and the matrix technology were used to mine the frequent pattern of new frequent tree.According to the frequent pattern mining of data flow,different data flow was classified into corresponding topic groups.Based on frequent item sets,the degree of similarity of multi-user access was calculated and the average access data of nearest neighbors to was found at the same time.Combined with topic extraction,nearest neighbor access and multi-user common interest similarity calculation,the data flow that best met user’s need was recommended.Thus,the hybrid recommendation for user data flow was achieved.Simulation results show that the proposed method has good safety performance of recommendation process and high accuracy of recommendation result.This method is a feasible method of data recommendation.