一种基于网站聚合和语义知识的电影推荐方法

A Film Recommendation Method Based on Website Aggregation and Semantic Knowledge

针对传统个性化推荐方法中存在的稀疏性、冷启动、过度专业化且准确率低等问题,提出一种基于网站聚合和知识的电影推荐方法。利用网络爬虫聚合源网站对某部电影的相关推荐,得到待推荐电影集,使用电影知识构建基于本体论的电影模型,并在该模型的基础上给出一种学习用户偏好权重的算法,采用SimRank算法和加权平均值计算电影相似度,根据相似度高低向用户进行推荐。实验结果证明,该方法的推荐准确度在非实时推荐场景下较现有方法提高10%以上,且实时推荐的推荐质量有明显提高,在一定程度上解决了稀疏性、冷启动及过度专业化等问题。

To solve the shortcomings in traditional methods of personalized recommendation such as sparsity,cold-start,overspecialization and low accuracy problem,this paper proposes a recommendation method based on Website aggregation and knowledge.It gets a movie set to be recommended by Web crawler aggregating Websites,and also builds an ontologybased film model based on which that proposes an algorithm for learning the weights of user preference.It measures the similarity between movies using SimRank method and the weighted average to recommend to users according to the level of similarity.Experimental results show that the accuracy of this method is improved by about ten percent than the existing methods when it is used on non-real-time recommendation.And quality of recommendations is improved significantly on real-time recommendation.In some extent,sparsity,cold-start,overspecialization problem can be solved.