支持推荐非空率的关联规则推荐算法

发布时间：2018-05-01 19:02

本文选题：关联规则 + 推荐系统　；参考：《通信学报》2017年10期

【摘要】：现有的关联规则推荐技术在数据提取时主要侧重于关联规则的提取效率,缺乏对冷、热门数据推荐平衡性的考虑和有效处理。为了提高个性化推荐效率和推荐质量,平衡冷门与热门数据推荐权重,对关联规则的Apriori算法频繁项集挖掘问题进行了重新评估和分析,定义了新的测评指标推荐非空率以及k前项频繁项集关联规则的概念,设计了基于k前项频繁项集的剪枝方法,提出了优化Apriori算法且适合不同测评标准值的k前项频繁项集挖掘算法,降低频繁项集提取的时间复杂度。理论分析比较与实验表明,k前项剪枝方法提高了频繁项集的提取效率,拥有较高的推荐非空率、调和平均值和推荐准确率,有效地平衡了冷、热门数据的推荐权重。
[Abstract]:The existing association rule recommendation techniques mainly focus on the efficiency of association rule extraction when data extraction, lack of cold, hot data recommendation balance consideration and effective processing. In order to improve the efficiency and quality of personalized recommendation and to balance the recommended weight between the unpopular and popular data, the frequent itemset mining problem of Apriori algorithm for association rules is reevaluated and analyzed. In this paper, we define the concepts of new evaluation index recommendation non-empty ratio and frequent itemsets association rules of k-front terms, and design a pruning method based on k-preterm frequent itemsets. An algorithm for mining k-front frequent itemsets with optimized Apriori algorithm and suitable for different evaluation standard values is proposed to reduce the time complexity of frequent itemsets extraction. The theoretical analysis and experiment show that the pruning method improves the efficiency of extracting frequent itemsets, has a high recommendation non-empty rate, harmonizes the average value and the recommendation accuracy, and effectively balances the recommended weights of cold and hot data.
【作者单位】：北京工业大学信息学部计算机学院;国网英大国际控股集团有限公司信息化工作部;
【基金】：国家自然科学基金资助项目(No.91646201,No.91546111) 北京市自然科学基金资助项目(No.4153058,No.4113076) 北京市教委面上基金资助项目(No.KM201710005023)~~
【分类号】：TP391.3

【相似文献】