Abstract: In order to improve the accuracy of forecast for new-user service, this paper proposes a collaborative filtering algorithm fusing multi-system user information. The algorithm finds neighbor and recommendation item by fusing multi-system user information to low dimensional manifold. It introduces the application process of manifold learning algorithm in recommendation service. Result shows that the proved algorithm can be more effective and accurate than traditional collaborative filtering algorithms for new user by contrastive experiment.

Key words: local invariable projection, collaborative filtering, new user

摘要： 为了提高新用户服务的预测准确率，提出一种融合多系统用户信息的协同过滤算法。该算法通过将多个系统的用户信息融合到低维流形中为用户寻找邻居和推荐项目，并介绍流形学习算法在推荐服务中的应用过程。通过对比实验，结果表明该算法比传统协同过滤算法能更有效、准确地为新用户提供推荐。

关键词: 局部不变投影, 协同过滤, 新用户

CLC Number: 

• TP301.6