提出了一种受非负条件约束、基于不完全矩阵隐特征分析的协同过滤模型，用以在非负条件约束下，预测未知的Web服务QoS（quality-of-service）数据。其主要原理是，将非负矩阵因式分解模型中基于矩阵操作的非负更新规则，扩展为只依赖于单一隐特征的非负更新规则，以获得仅依赖于所需模型中单一隐特征、以及目标稀疏矩阵中已知元素的非负训练过程。为提高最终模型的预测精度，模型中进一步引入了Tikhonov 正则化，并对隐特征的非负更新规则进行了相应扩展。在2个大型Web服务QoS数据集上的实验表明，该模型能在满足非负条件约束的前提下，准确地预测未知的QoS数据。%Based on LF-analysis on incomplete matrices ,a non-negative collaborative filtering model is proposed for predicting un-known QoS data under the non-negativity constraints in this paper .The principle of this model is to transform the traditional ma-trix manipulation-based non-negative update rule into single-element-dependent non-negative update rule ,in order to obtain a non-negative training process relying on each single LF of the desired model and the known data of the target matrix .To improve the prediction accuracy of the resulting model ,the Tikhonov regularization is integrated into the model .Experimental results on two QoS datasets demonstrate that the proposed model is highly effective in predicting unknown QoS data .
展开▼
