Service Classification Based on Improved BP Neural Network

Abstract

With the development of the Internet, several candidate services have emerged for achieving the same task, most of which are functionally identical but different in non-functional properties. Therefore, these services can be classified into different service-quality levels. The so-called Quality of Service (QoS) comprises a set of non-functional properties that can be used to efficiently classify and rank these various services. In this paper, an algorithm called CNBP is proposed to address the problem of automatically classifying services. The core idea of this algorithm is that the weights and biases of a back-propagation network are optimized by a hybrid optimization based on two algorithms: the Nelder-Mead simplex algorithm and the Cuckoo search algorithm. The improved back-propagation (BP) classifier is used to classify candidate services into different QoS levels. Through experiments based on the Quality of Web Services dataset and a comparative analysis with traditional back-propagation networks and three other classification algorithms, we demonstrate that the proposed algorithm performs well in terms of its classification accuracy and stability.