Traditional collaborative filtering recommendation algorithms have the problem of over dependence on neighbouring users in the prediction of target users’rating while neglecting the rating characteristic of target users themselves.Aiming at the problem,this paper proposes an improved RBF-based neural network prediction method.The method uses RBF (radial basis function)neural network to carry out model training on the projects rating data of neighbouring users first,and gets the target user-based network rating model;then it calculates in combination with the rating of target user its own to obtain a rating result which is based on the model;at last,it combines the model ratings of all the neighbouring users to predict the final rating of target user on target items.The improved algorithm learns from the similarity between neighbouring users and considers the rating characteristics of target user its own as well.Experimental results show that the improved collaborative filtering recommendation algorithm is able to achieve better recommendation results than the traditional algorithms.%针对传统协同过滤推荐算法对目标用户的评分预测过于依赖邻近用户，而忽略目标用户自身评分特性的问题，提出一种改进的基于径向基 RBF（Radial Basis Function）神经网络的预测方法。该方法首先使用 RBF 神经网络对邻近用户的项目评分数据进行模型训练，得到基于该用户的网络评分模型；然后结合目标用户自身的评分进行计算，得到一个基于该模型的评分；最后结合所有邻近用户的模型评分预测出目标用户对目标项目的最终评分。改进后的算法既借鉴了用户之间的相似性，也考虑了目标用户自身的评分特性。实验结果表明，改进后的协同过滤推荐算法可以获得比传统算法更好的推荐效果。
展开▼
