Colbar: A collaborative location-based regularization framework for QoS prediction
Colbar: A collaborative location-based regularization framework for QoS prediction
复制标题
Colbar：用于 QoS 预测的基于位置的协作正则化框架
DOI：
10.1016/j.ins.2013.12.007
复制
发表时间：
2014-05-01
影响因子：
8.1
通讯作者：
Xiong, Naixue
中科院分区：
文献类型：
--
作者：
Yin, Jianwei;Lo, Wei;Xiong, Naixue
服务质量(Qos)是面向服务计算(SOC)领域的一个基本要素。在Web 2.0时代，预测缺失的QOS值变得越来越重要，因为它是许多面向服务的应用程序不可或缺的前期处理。以前的研究低估了用户地理信息的重要性，我们认为这将有助于提高Web服务调用过程中的预测精度。本文提出了一种新的基于位置的协作正则化框架(COLBAR)来解决个性化服务质量预测问题。我们首先利用个人地理和服务质量信息来确定可靠的社区。然后，我们集思广益，构造了两个基于位置的正则化项，并将这两个项集成在一起，建立了一个统一的矩阵分解框架。最后，我们进行中间融合以产生更好的预测结果。在大规模的真实服务质量数据集上的实验分析表明，COLBAR的预测精度在各种标准上都优于其他最先进的方法。(C)Elsevier Inc.出版的2013年。
Quality-of-Service (QoS) is a fundamental element in Service-Oriented Computing (SOC) domain. At the ongoing age of Web 2.0, predicting the missing QoS values becomes more and more important since it is an indispensable preprocess of numerous service-oriented applications. Previous research works on this task underestimate the importance of users' geographical information, which we argue would contribute to improving prediction accuracy in Web services invocation process. In this paper, we propose a novel collaborative location-based regularization framework (Colbar) to address the problem of personalized QoS prediction. We first leverage the personal geographical and QoS information to identify robust neighborhoods. And then, we collect the wisdom of crowds to construct two location-based regularization terms, which are integrated to build up an unified Matrix Factorization framework. Finally we make intermediate fusions to generate better prediction results. The experimental analysis on a large-scale real-world QoS dataset shows that the prediction accuracy of Colbar outperforms other state-of-the-art approaches in various criteria. (C) 2013 Published by Elsevier Inc.