郭丽莎,邓棉予,李秋雨,冯琪,郭仲凯.基于协同过滤算法的电视产品打包推荐[J].中南民族大学学报自然科学版,2020,39(6):655-660
基于协同过滤算法的电视产品打包推荐
Package recommendation for TV products based on collaborative filtering algorithm
  
DOI:10.12130/znmdzk.20200618
中文关键词: 协同过滤算法  打包推荐  推荐指数  查准率  查全率
英文关键词: collaborative filtering algorithm  package recommendation  recommendation index  precision  recall
基金项目:国家自然科学基金资助项目(11801575);中央高校基本科研业务费专项资金资助项目(CZQ20008)
作者单位
郭丽莎 中南民族大学 数学与统计学学院 武汉 430074 
邓棉予 中南民族大学 数学与统计学学院 武汉 430074 
李秋雨 中南民族大学 数学与统计学学院 武汉 430074 
冯琪 中南民族大学 数学与统计学学院 武汉 430074 
郭仲凯 中南民族大学 数学与统计学学院 武汉 430074 
摘要点击次数: 204
全文下载次数: 129
中文摘要:
      通过研究用户和产品之间的关系,解决电视产品精准营销推荐和打包推荐两个问题.借鉴基于产品的协同过滤算法,用户与产品相关度为推荐系数,分析其关联关系;通过网络爬虫获取辅助信息对用户和产品贴多级标签,按标签对打包用户推荐打包产品.在IBCF、random、popular三种算法中,选择popular算法时,每个用户最优推荐产品个数为15个;选择IBCF算法时,每类用户的打包推荐产品包个数为3个,分别得到推荐方案.在查准率、查全率以及综合评价指标三个评价指标数据的对比中,打包推荐的3项评价指标明显优于精准推荐,其中查准率的值大幅提高,综合评价指标的值增长两倍,查全率的值上调.
英文摘要:
      By studying the relationship between users and products, two problems of accurate marketing recommendation and packaging recommendation for TV products are solved. This paper uses the product-based collaborative filtering algorithm and sets the user-product relevance as the recommendation coefficient to analyze its association. The auxiliary information obtained by the web crawler attaches a multi-level label to the user and the product, and recommends the packaged product to the packaged user according to the labels. Among the three algorithms of IBCF, random, and popular, the number of optimal recommended products per user is 15 by selecting the popular algorithm. The number of package recommended products for each type of user is 3, which is recommended by the IBCF algorithm. Comparing three evaluation indexes, namely, precision, recall and comprehensive evaluation index, the three evaluation indexes of packaged recommendation are obviously better than precise recommendation. Among them, the value of precision is greatly increased, the value of comprehensive evaluation index is doubled, and the value of recall is increased.
查看全文   查看/发表评论  下载PDF阅读器
关闭