欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
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。
英文摘要：
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.
同期刊论文项目
同项目期刊论文
Phase diagrams on an unsignalized intersection for the cases of different maximum velocities of vehi
Stochastic resonance in an optical bistable system subjected to cross-correlated additive white nois
SPREAD OF EPIDEMIC ON COMPLEX NETWORKS UNDER VOLUNTARY VACCINATION MECHANISM (vol 21, pg 1197, 2010)
Investigation on Relationship between Lubricating Performance and Stability of Emulsion for Cold Str
期刊信息
位置：