一种多层混合的推荐模型研究

Research of a multi-layer hybrid recommend model

在全面分析研究邻域模型和隐语义模型的基础上,针对现有推荐模型的优缺点,首先建立了一种融合邻域模型和隐语义模型的混合Top-N推荐模型,有效利用了用户反馈的信息,并全局考虑了用户与项目的潜在关系。然后综合考虑用户反馈信息、自身特征信息及潜在信息等因素,提出了一种基于SVD++上的全新混合Top-N推荐模型SHT(基于SVD++的混合Hybrid Top-N推荐模型缩写)。通过将特征信息融入模型,可准确地表现用户与项目的属性特征,实现依照用户的喜好与习惯高效、快捷和精准地推荐。实验结果表明,每一层的模型都能够在不同程度上提高推荐结果的精度。

Comprehensively analyzing the neighborhood model and the lingo model,we summarize the characteristics of existing technologies as well as the advantages and disadvantages of them.A hybrid Top-N model is established in order to effectively use the user feedback information and generally consider the potential relationship between users and items,by fusing the neighbor-hood model and lingo.With comprehensive consideration on the user feedback information,characteristic information,and poten-tial information,we propose a novel multi-tier hybrid Top-N model based on SVD++,called SHT (SVD++-based Hybrid Top-N Recommender Model).It is able to recommend the most accurate and effective items to user by their preferences and hab-its.The experimental results show that each layer of the model can improve the accuracy of recommendation results in different extents.