基于Softmax回归模型的协同过滤算法研究与应用  

Research and Application of Collaborative Filtering Algorithm Based on Softmax Regression Model

作　　者：孟佩[1] 曹菡[1] 师军[1] 

机构地区：[1]陕西师范大学,陕西西安710119

出　　处：《计算机技术与发展》2016年第12期153-155,159,共4页Computer Technology and Development

基　　金：国家自然科学基金资助项目(41271387)

摘　　要：针对传统的协同过滤推荐算法所存在的推荐精度不高的问题,提出了基于Softmax回归模型的协同过滤算法。根据用户的属性特征将用户分为不同的簇,再从目标用户所在的簇中实现协同过滤推荐,有效缩减了最近邻居的查找范围,提高了推荐效率。主要将此改进算法应用于饮食推荐中,根据用户的饮食记录对用户按口味偏好进行精准分类,将偏好相同的用户划分到同一个簇中,再从目标用户所在的用户簇中查找最近邻居,完成推荐。从两方面对此方法进行了实证分析:基于Softmax的用户口味偏好分类的准确率分析和基于Softmax的协同过滤推荐精准度分析,验证了该方法的有效性和可行性。In view of the low accuracy for traditional collaborative filtering recommendation algorithm,the collaborative filtering algorithm based on Softmax regression model is proposed. According to the user＇ s attributes, the users can be divided into different clusters, and the collaborative filtering recommendation is realized in the cluster from its target users,reduction of the nearest neighbors search scope,improvement of the performance of the recommendation system. The improved algorithm is applied to dietary recommendations,depending on the user＇ s diet by recording the user taste preferences for accurate classification, the same user preferences will be divided into the same cluster, and then the nearest neighbor is searched from the user cluster where there is the target user to complete the recommendation. An empirical analysis about this method from two aspects is made,including the accuracy analysis of the user＇ s taste preference classification based on Softmax and precision analysis of collaborative filtering recommendation based on Softmax, and the effectiveness and feasibility is verified.

关 键 词：Softmax回归 口味偏好 协同过滤 营养饮食 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]