关联规则和聚类分析在个性化推荐中的应用

Application of Association Rules and Clustering Analysis to Personalized Recommendation

提出了两种应用访问页面关联规则和访问模式聚类分析结果相结合进行个性化推荐的方法,即将聚类分析作为关联规则的预处理和将关联规则和聚类分析互补使用,并与单独应用访问页面关联规则或访问模式聚类分析结果进行个性化推荐时的推荐测度进行了比较·实验表明,将聚类分析作为关联规则的预处理的推荐方法可以显著地提高推荐的准确率,而将关联规则和聚类分析互补使用的推荐方法具有较高的推荐覆盖率·同时发现将聚类分析和关联规则结合使用并不能同时改善推荐的准确率和覆盖率·

?Combining the association rules discovered from pageviews with clustering analysis of access patterns, two methods are presented for personalized recommendation. The first is to take the clustering analysis as a pre-processing procedure for association rules discovery, and the second is to have the association rules supplement clustering analysis. Then, a comparison of the measure of recommendation is made between the two methods and using separately the association rules from pageviews or the clustering analysis of access patterns. The result of comparative testing shows that the first method can improve greatly the precision rate of recommendation, whereas the second one has comparatively high coverage rate of recommendation. In addition, it is found that combining the association rules with clustering analysis is unable to improve both the measure precision and coverage rate of recommendation simultaneously.