摘要
目前推荐系统中协同过滤算法是应用最成熟的推荐算法之一,然而传统算法没有考虑随着时间的迁移,用户的兴趣也可能发生相应变化以及特征属性在推荐过程中对推荐结果的影响,致使预测结果不准确。为此,提出一种新的相似性改进算法对传统算法进行改进。改进后的协同过滤算法对基于时间的Logistic权重函数与用户特征属性进行加权计算,形成一种新的相似性度量模型。实验结果表明该算法推荐平均绝对误差(MAE)比传统算法降低了12%,较传统算法推荐质量有明显提高。
At present,collaborative filtering algorithm is one of the most mature recommendation algorithms applied in recommendation systems. However,traditional collaborative filtering algorithms do not take into account the problem of users' interests drifting over time as well as the effects of feature attributes,which may decrease the accuracy of recommendation results. Hence,in order to enhance the traditional algorithms,a novel similarity measurement algorithm is put forward. In this paper,an innovative similarity measurement model is constructed by combining time-based Logistic weight function and user feature similarity-based data weight. Experimental results show that compared with traditional algorithms,the mean absolute error(MAE) of recommendation using the proposed algorithm is reduced by an average of 12% and the quality of recommendation is improved significantly.
出处
《计算机应用与软件》
2017年第2期285-289,312,共6页
Computer Applications and Software
基金
河南省科技攻关项目(142402210435)
关键词
协同过滤
兴趣变化
时间权重
用户特征
Collaborative filtering
Interest change
Time weight
User feature