摘要
为解决推荐算法中新物品完全冷启动问题,针对数据精准性不足和用户个性化缺失问题,提出一种完全冷启动个性化推荐算法。在运用过滤方法形成精准数据的基础上,引入个性化因子,改变原有物品的相似关联性,推荐依据会随着用户个性化特征而有所变动。经过对比分析,发现融入个性化的完全冷启动推荐算法仅查全率基本不变,精确率、假正率和F1值多个评价指标得到提升,此外,P-R曲线、ROC曲线以及提升曲线都说明该算法具有更好的推荐效果。
To solve the problem of complete cold start on new items in recommendation algorithm,a complete cold start personalized recommendation algorithm was proposed in view of the lack of data accuracy and user personalization.With accurate data formed using filtering method,the personalized factor was introduced to adjust the original items’similarities,thus recommending basis change with the user’s personalized characters.Through the analysis of comparison,the personalized recommendation algorithm for completely cold start is found with the almost unchanged recall rate,as well as the improved precision rate,false positive rate and F1 value.In addition,P-R curve,ROC curve and depth-lift curve also show that the algorithm has better recommendation results.
作者
李剑锋
陈海龙
翟军
林岩
LI Jian-feng;CHEN Hai-long;ZHAI Jun;LIN Yan(Department of Manage Science and Engineering,School of Maritime Economics and Management,Dalian Maritime University,Dalian 116026,China)
出处
《计算机工程与设计》
北大核心
2024年第8期2329-2335,共7页
Computer Engineering and Design
基金
国家自然科学基金项目(72271037)
中央高校基本科研业务费专项资金基金项目(3132019353)。
关键词
推荐算法
完全冷启动
个性化推荐
近相邻算法
物品冷启动
过滤方法
数据精准性
recommendation algorithm
completely cold start
personalized recommendation
neighbor algorithm
item cold start
filtering method
data accuracy
