面向协同过滤推荐的多粒度用户偏好挖掘研究

Research on Multi-granularity Users' Preference Mining Based on Collaborative Filtering Personalized Recommendation

【目的】针对协同过滤中用户偏好挖掘粒度与挖掘效率之间的关系展开研究,以期找出效率最高的挖掘粒度。【方法】结合实际应用情况将用户偏好挖掘粒度从粗到细划分为三种,并对三种粒度下相应的偏好挖掘算法进行详细设计,通过实验对比不同粒度下用户偏好挖掘的效率。【结果】实验结果表明,当用户偏好挖掘粒度从粗到细变化时,偏好挖掘效率也会逐渐降低。【局限】以用户消费及评分数据为挖掘用户偏好的数据来源,对于其他类型数据源暂未涉及。【结论】粗粒度的偏好挖掘能更好地发现用户偏好。

[Objective] Researching the relationship between users＇ preference mining granularity and mining efficiency in collaborative filtering, this paper aims at finding out the most efficient mining granularity. [Methods] According to the practical application, the users＇ preference mining granularity is divided into three kinds from coarse-grained to fine-grained, and then design the corresponding preference mining algorithm under the three kinds of granularities, finally contrast users＇ preference mining efficiency under different granularities through experiments. [Results] Experimental results show that the preference mining efficiency reduces as the users＇ preference mining granularity changes from coarse to fine. [Limitations] Data only includes users＇ consumption data and rating data, other types of data are not covered temporarily. [Conclusions] Coarse-grained preference mining is better for discovering users＇ preferences.