摘要
为进一步提高推荐系统的推荐精度,提出一种新的基于项目云的有序秩聚类协同过滤推荐算法,其中包括三大步:数据处理,有序聚类,生成推荐。该方法不仅借助定性分析思想利用项目云有效地填充了缺失数据,而且通过对项目分布的数字特征做排序、分割、聚类,在类内产生"邻居",大大缩短了计算时间。通过在MovieLens数据集上的实验表明,在平均绝对误差和预测精确度上,该算法确实优于传统推荐算法。
In order to further improve recommender accuracy,in this paper we propose a novel ordered rank cluster in collaborative filtering based on the item cloud(ICORC)method,which includes three steps:data processing,ordered rank clustering,and recommendation generating.This method has two advantages.One is that it can tackle the data sparsity problem by filling in missing data using the item cloud.Another distinct feature is that it can save time and obtain more accuracy through finding"neighbors"of items among the clusters,which are formed by sorting,partition and clustering for the numerical characteristics of item distribution.To the best of our knowledge,there has been no prior work on investigating CF recommendation by combining ordered rank cluster.We conducted this experiment on the MovieLens datasets and found that ICORC is superior to other collaborative filtering(CF)algorithms on the mean absolute error and Precision.
出处
《太原理工大学学报》
CAS
北大核心
2016年第5期673-679,共7页
Journal of Taiyuan University of Technology
基金
国家自然科学基金资助项目:高维数据变量间非线性交互作用的研究(11571009)
关键词
协同过滤
云模型
有序秩聚类
评分可靠度
推荐系统
collaborative filtering
cloud model
ordered rank cluster
rating reliability
recommender system