摘要
推荐系统作为帮助用户快速获取有用信息的重要工具之一,得到了深入的研究和广泛的应用。为了更好地挖掘基于上下文信息的推荐场景中“用户-项目-上下文”三者之间的潜在关系,定位用户可能的兴趣点,进而提升推荐准确度和用户体验。通过分析目前已有的关于上下文情景数据处理技术的原理和特点,在时间复杂度或者时效性方面仍存在着不足之处,这也是目前融合上下文推荐算法的研究重点和难点之一。针对推荐的时间复杂度高这一方面,提出了融合因子分解技术的推荐算法(FM-CR)来提升基于上下文推荐的效果,包括用户接受率和时间复杂度。经过实验证明,提出的融合因子分解技术的推荐算法(FM-CR)能够在保证用户的推荐接受率的情况下,使得产生推荐结果的时间效率有了明显提高。
As one of the important ways to help users get information what they need quickly and efficiently,recommender sys⁃tem has been deeply studied and widely applied.To better explore potential relationship among"user-item-context"and locate the interest points of users,which will help improve the accuracy of recommendation and user experience.By analyzing the principles and characteristics of context-information processing technology,it is found that deficiencies in time complexity or timeliness are still existed,which is the hot research issue of context recommendation algorithms.In view of the high time complexity,a recommen⁃dation algorithm based on factorization machines,named as FM-CR is proposed.The FM-CR algorithm tries to improve the recom⁃mended effect,including user acceptance rate and time complexity.The experimental results show that FM-CR algorithm can signif⁃icantly increase the time efficiency while it keeps the considerable recommendation-acceptance rate.
作者
乔雨
圣文顺
QIAO Yu;SHENG Wenshun(Nanjing Tech University Pujiang Institute,Nanjing 211200)
出处
《计算机与数字工程》
2021年第6期1123-1126,1204,共5页
Computer & Digital Engineering
基金
江苏省高校自然科学研究面上项目(编号:19KJD520005)资助。
关键词
推荐系统
上下文信息
因子分解
时间复杂度
推荐接受率
recommender system
context information
factorization machines
time complexity
recommendation-accep⁃tance rate
