融合情境信息的非支配排序多目标进化推荐算法  

A Multi-Objective Evolutionary Algorithm for Recommendation Using Non-Dominated Sorting and Contextual Information

作　　者：朱鑫 金友振 夏小云 Zhu Xin;Jin Youzhen;Xia Xiaoyun(College of Mathematics and Computer Science,Zhejiang Normal University,Jinhua,Zhejiang 321004;College of Information Science and Engineering,Jiaxing University,Jiaxing,Zhejiang 314001)

机构地区：[1]浙江师范大学数学与计算机科学学院,浙江金华321004 [2]嘉兴学院信息科学与工程学院,浙江嘉兴314001

出　　处：《嘉兴学院学报》2023年第6期79-85,共7页Journal of Jiaxing University

基　　金：浙江省普通本科高校“十四五”教学改革项目(jg20220434);浙江省公益技术应用研究计划项目(LGG19F030010);嘉兴学院实践教学专项改革研究项目(SJZY20072307-007);嘉兴学院课程思政示范基层教学组织项目(851521077);嘉兴学院“勤慎青年学者”培养计划资助(嘉院人字〔2023〕12号)。

摘　　要：为了解决传统的推荐技术存在的推荐列表精确度和多样性冲突等问题,提出了一种融合情境信息的基于非支配排序多目标进化算法求解方法.使用K-means++聚类技术对用户进行聚类操作,按照类间用户差异性最大、类内用户相似性最大原理划分族群.结合协同过滤算法融入时间情境信息追踪用户偏好漂移,并通过非支配排序进化算法NSGA-II平衡推荐列表的准确度和多样性两个指标.在公共数据集Movielens-1M上将提出的非支配排序多目标进化情境推荐算法与现有算法进行对比,实验结果表明,该算法减少了预测评分时的平均绝对误差值,在准确度和多样性两个评价指标上均有提高.In order to solve the problems such as the conflict between the accuracy and the diversity of a recommendation list in traditional recommendation technology,this paper proposes a recommendation algorithm based on a multi-objective evolutionary algorithm which uses contextual information and non-dominated sorting.K-means++clustering technology is used to cluster users,and groups are divided according to the principle of maximum difference between classes and maximum similarity among users within a class.Combining the collaborative filtering algorithm integrated with time context information,the preference drift of users is tracked,and the accuracy and diversity of a recommendation list are balanced by the evolutionary algorithm NSGA-II using the non-dominated sorting.On the public dataset Movielens-1M,the multi-objective evolutionary recommendation algorithm using the non-dominated sorting and contextual information proposed in this paper is compared with existing recommendation algorithms.The results show that the proposed algorithm is better than the ones compared in terms of the accuracy and diversity of the evaluation index of the recommendation list,which reduces the mean absolute error of the predicted score.

关 键 词：融合情境信息 多目标进化算法 推荐算法 情境信息 聚类技术 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]