基于时间因子改进个性化推荐模型
2021-08-05胡安明陈惠娥
胡安明 陈惠娥


摘 要:传统推荐系统算法模型主要集中研究用户偏好与物品的关联性,根据用户主观意见进行推荐,未充分考虑用户与物品所处的客观环境,造成推荐时的实际偏差。本文基于传统推荐算法引入时间因子,提高模型推荐效果。实现方法主要是通过比较引入与未引入时间因子,使用UserCF算法和ItemCF算法观察MAE值的大小变化情况。时间因子的引入,改善了传统推荐系统算法模型主要集中研究用户偏好与物品的关联性等方面的推荐失真问题,提高了模型推荐的可靠性和实用性。实验结果表明,引入时间因子能对传统协同过滤算法在MAE指标方面有一定提高,计算效果优于传统推荐算法。
关键词:时间因子;个性化推荐;协同过滤
中图分类号:TP311.60 文献标识码:A
Improved Personalized Recommendation Model based on Time Factor
HU Anming1, CHEN Huie2
(1.School of Computer Science and Engineering, Guangzhou Institute of Technology, Guangzhou 510540, China;
2.Guangdong University of Finance, Guangzhou 510521, China)
anminghu@qq.com; 318802207@qq.com
Abstract: Traditional model of recommendation system algorithm mainly focuses on the relationship between user preferences and items, and makes recommendations according to users' subjective opinions. It fails to take into full consideration the objective environment of the user and the item, resulting in actual deviation in recommendation. This paper proposes to improve model recommendation effect by introducing time factor into traditional recommendation algorithm. The improved model is realized by comparing algorithms with and without time factor, and using UserCF algorithm and ItemCF algorithm to observe the changes of MAE (Mean Absolute Error) values. Introduction of time factor improves the algorithm model of traditional recommendation system, which mainly focuses on the recommendation distortion of user preferences and the relevance of items, so to improve the reliability and practicability of the model recommendation. Experimental results show that introduction of time factor can improve MAE index of traditional collaborative filtering algorithm, and calculation effect is better than that of traditional recommendation algorithm.
Keywords: time factor; personalized recommendation; collaborative filtering
1 引言(Introduction)
近年来,随着计算机技术的发展和网络的普及,大量数据信息融入互联网。面对如此巨量的数据信息资源,如何让用户对其进行精确高效的查询,有效使用互联网资源;如何根据用户的个人偏好信息,结合客观环境,有效地处理推荐用户所需的信息资源,仍是目前推荐系統研究的热点。用户—物品间的浏览访问记录和上下文辅助信息数据,为推荐系统提供了数据挖掘分析的基础[1],如何从这些复杂的数据中挖掘出准确的用户偏好信息,也是当前研究的热点。
传统的推荐系统模型主要关注用户与物品的行为数据研究,根据用户的行为数据,挖掘出用户的特征偏好,从而进行个人用户的推荐。……
