摘要
为了提高电影个性化推荐的准确性,将电影通过导演、演员、上映时间、类型和地区等五个部分作为特征维度来表征,特征维度权值采用CHI方法计算,特征维度的权值进行归一化后,电影之间的相似度可以通过特征维度间的相似度体现,用户推荐模型通过不断迭代更新对各维度特征权值进行修正,提高模型推荐的准确性。实验结果表明,改进的算法在MovieLens数据集能够获得较高的准确率和召回率,能够比较准确地捕获用户的兴趣,并在一定程度上解决了用户兴趣漂移的问题。
In order to improve the accuracy of the movie personalized recommendation,the movie is charactered with thefeature dimensions of director,performer,showtime,type and area. The weights of feature dimensions are calculated with CHImethod,and then normalized to reflect the similarity among movies by means of the similarity among the feature dimensions.The user recommendation model can correct the feature weigh of each dimension with the continuous iteration update to improvethe accuracy of model recommendation. The experimental results show that the improved algorithm can obtain high accuracy rateand recall rate with MovieLens dataset,capture the user′s interest exactly,and solve the user interest drifting to a certain extent.
作者
吴承毅
WU Chengyi(Southwest University of Science and Technology,Mianyang 621010,China;Shangluo Vocational & Technical College,Shangluo 726000,China)
出处
《现代电子技术》
北大核心
2016年第15期127-129,共3页
Modern Electronics Technique
关键词
多维度
电影推荐
权值动态更新
个性化推荐模型
multi.dimension
movie recommendation
weight dynamic update
personalized recommendation model
