摘要
推荐算法是个性化推荐系统中最为核心的一部分。文本通过给出产品流行性定义,提出了一种改进的用户兴趣点度量方法,进而将用户的兴趣点嵌入到基于物质扩散原理的推荐算法中。新算法引入参数口度量产品的推荐权重与用户兴趣点之间的关系。MovieLens数据集上的数值结果表明新的用户兴趣点定义方法可以同时改进推荐算法的准确度和推荐列表多样性,当采用60%的数据作为训练集时,多样性可以提高13.15%。进一步的结果表明当训练集很稀疏的时候,应当赋予与用户兴趣点不同的产品更高的推荐能力,随着稀疏度增加,赋予与用户兴趣点相近的产品更多推荐能力可以大幅度提高算法的表现。
Recommendation algorithm plays core role in recommendation systems. By introducing a new object popularity definition, this paper presents an improved user taste measurement, which is embedded into the mass-diffusionbased algorithm. In the new algorithm, a free parameter β is introduced to investigate the correlation between the object popularity and user tastes. The numerical results on one benchmark dataset show that both the accuracy and diversity could be improved greatly. For example, when the training set is set as 60% percentage of data, the diversity could be enhanced 13.15%. In addition, the numerical results indicate that the objects whose popularities are far away to users' tastes should be set more recommendation power when the training data is sparse, on the contrary, the recommendation power of objects whose degrees are close to the target user' s tastes should be enhanced when the sparsity of training data is increased.
出处
《情报学报》
CSSCI
北大核心
2011年第11期1158-1162,共5页
Journal of the China Society for Scientific and Technical Information
基金
基金项目:国家自然科学基金资助项目(10905052,70901010,71071098,91024026)
上海市科研创新基金(11ZZ135,11YZ110)
上海市智能信息处理重点实验室开放基金(IIPL-2010-06)
上海市大学生创新项目(SH091025227)
上海市系统分析与集成重点学科(S30501).
关键词
推荐算法
用户兴趣点
物质扩散
二部分网络
recommendation algorithm, user tastes, mass diffusion, bipartite networks