Cuckoo Optimized Multi Layer Perceptron for Web-Services Classification

Abstract

Web services describe a collection of operations that are network-accessible through standardized web protocols and its features are described by using a standard eXtensible Markup Language (XML)-based language. Web Services allows integration of applications with ease and less cost. Web Services faces the problem of classifying the categories from a predefined set. In this paper, the focus is on the classification of Web services for managing using Multi-Layer Perceptron-Neural Network (MLP-NN). The MLPNNs are universal approximators and its performance is dependent on the number of hidden neurons. The training accuracy could also be affected by several other parameters, including the number of layers, the number of training samples, the length of learning period, the choice of neuron activation functions, and the training algorithm. To optimize the number of hidden neurons, numerous techniques have been established in literature, which associate it input and output layer sizes or with the number of training samples. A Cuckoo Search (CS) algorithm is proposed to optimize the structure of the MLPNN for improving web service classification.

Keywords: Cuckoo Search (CS), Multi-Layer Perceptron-Neural Network (MLP-NN), Web services, QWS dataset.

View PDF