针对电子商务中多准则决策协同过滤推荐系统的稀疏性问题,提出了一种基于神经网络的多准则决策协同过滤推荐系统,设计了多个相似性度量混合的机制来缓解稀疏性问题.首先,采用领域本体提取选项的信息,并且提取用户的访问信息;然后,使用模糊理论对用户的评价信息进行建模,解决用户评价的不确定性与复杂性;最终,采用自适应神经模糊推理系统预测选项的总评分与选项各个准则评分之间的关系,设计了基于梯度下降法的凸组合机制,提高了冷启动用户相似性度量的鲁棒性与可靠性,最小化系统的预测误差.实验结果显示本算法有效地缓解了稀疏性问题,并且获得了较高的推荐准确率.“,”In order to reduce the sparsity problem of collaborative filtering recommendation based on multi-criteria decision making in e-commerce, an enhanced collaborative filtering recommendation based on multi-criteria decision making and neural networks is proposed, and a mechanism of multi-similarity measure mixture is designed to reduce the problem of sparsity. Firstly, the domain ontology is adopted to abstract the information of items, and the information of users is abstracted too; then, the user ratings and comments are modeled by fuzzy theory to solve the uncertainty and complexity of the user ratings and comments; lastly, adaptive network based fuzzy inference system is used to predict the relationship between the total ratings and the ratings of each criteria, the convex combination mechanism based on the gradient descent method is designed to improve the robustness and reliability of the similarity measure of cold start users, at the same time, the prediction error of the recommendation system is minimized. The experimental results show that, the proposed algorithm reduces the sparsity problem efficiently, and performs a good recommendation accuracy.