摘要
模拟现实电子商务推荐场景,设计三支交互推荐模型,提出结合流行度区间和M-distance k近邻的混合推荐算法.在"推荐""不推荐"基础上引入"促销"构成三支,以丰富系统行为、降低推荐代价;构造序贯三支人机交互,持续学习用户的消费习惯,并提供更精准的推荐;根据目标用户的消费记录,选择适当的推荐策略.在多个数据集上的实验结果表明,与其他粗糙集模型下的算法相比,该算法的平均代价更低.
In order to simulate the real-world e-commerce recommendation scenario,this paper designs a three-way interactive recommendation model,and proposes a hybrid recommendation algorithm combining the popularity range and M-distance k-nearest neighbors.The proposed model introduces"promote"on the basis of"recommend"and"not recommend"to form the three-way,enriches the behavior of the system,and reduces the recommendation cost.The system constructs sequential three-way human-computer interactions,learns user preferences continuously and provides more accurate recommendations.It also selects the appropriate recommendation strategy based on the target user’s consumption record.Experimental results on several datasets show that the average cost of the proposed algorithm is lower than those based on other rough set models.
作者
徐媛媛
张恒汝
闵帆
黄雨婷
Xu Yuanyuan;Zhang Hengru;Min Fan;Huang Yuting(School of Computer Science,Southwest Petroleum University,Chengdu,610500,China;Institute for Artificial Intelligence,Southwest Petroleum University,Chengdu,610500,China)
出处
《南京大学学报（自然科学版）》
CAS
CSCD
北大核心
2019年第6期973-983,共11页
Journal of Nanjing University（Natural Science）
基金
国家自然科学基金(61802321,61379089)
关键词
交互推荐
促销代价
推荐系统
三支决策
interactive recommendation
promotion cost
recommender system
three-way decision
