Service recommendation aims to help users to find the most suitable Web services based on their quality of service (QoS) preferences instead of searching through extensive volume of Web services using search engine manually. Accurate unknown QoS rating prediction is one of the key challenges in the analysis of service recommendation. Collaborative filtering (CF) is a well-known recommendation method that estimates missing ratings by employing a set of similar users to the active user. The core idea of CF consists of picking out an appropriate set of users and using them in the rating prediction process. However, the majority of existing CF methods are not well-designed for Web service QoS prediction as they ignore the implicit but important characteristic of Web service QoS data that fluctuate in big-range. In other words, through analysis of real-world QoS datasets, we observed that QoS ratings vary widely and they are highly skewed with large variances, as two main facts, which dramatically degrade the accuracy of CF methods in QoS prediction. Towards this problem, in this paper, we propose a big-range aware collaborative filtering approach dubbed BRACF to predict Web service QoS ratings accurately. Specifically, since big-range of QoS data can lead to similarity exaggeration, we design a simple yet effective similarity model which considers the influence of big-range among users’ QoS data for accurately characterizing the similarity between users. Moreover, the similarity model is seamlessly incorporated into CF model for identifying similar neighbor using Top-K strategy and then it generates QoS predictions by combining bias information. Through extensive experiments on two public real-world Web service QoS for datasets, as response time and throughput, we show that BRACF significantly outperforms state-of-the-art CF methods. We believe that this work demonstrates the potential impact of big range data for the accurate QoS prediction.

中文翻译：

---

Web服务QoS预测：协作过滤遇到大范围数据波动时

服务推荐旨在帮助用户根据他们的服务质量（QoS）偏好找到最合适的Web服务，而不是使用搜索引擎手动搜索大量Web服务。准确的未知QoS等级预测是服务推荐分析中的关键挑战之一。协作过滤（CF）是一种众所周知的推荐方法，该方法通过使用一组与活动用户相似的用户来估计缺少的评分。CF的核心思想是挑选出一组合适的用户并将其用于收视率预测过程中。但是，大多数现有的CF方法并未为Web服务QoS预测精心设计，因为它们忽略了大范围波动的Web服务QoS数据的隐含但重要的特征。换一种说法，通过对现实世界中QoS数据集的分析，我们发现QoS等级差异很大，并且偏差大​​，且偏差大，这是两个主要事实，这大大降低了CF方法在QoS预测中的准确性。针对此问题，在本文中，我们提出了一种称为BRACF的大范围协作过滤方法，以准确预测Web服务QoS等级。具体来说，由于大范围的QoS数据可能导致相似度夸大，因此我们设计了一个简单而有效的相似度模型，该模型考虑了用户QoS数据之间大范围的影响，以准确地表征用户之间的相似度。此外，将相似度模型无缝整合到CF模型中，以使用Top-K策略识别相似邻居，然后通过组合偏差信息生成QoS预测。通过针对数据集的两个公共现实世界Web服务QoS（如响应时间和吞吐量）进行的广泛实验，我们表明BRACF明显优于最新的CF方法。我们相信这项工作证明了大范围数据对于准确QoS预测的潜在影响。