Transfer Function Modeling in Web Software Fault Prediction Implementing Pre-Whitening Technique

Abstract

The issue of Web reliability is gaining importance, as different Web-based applications are getting popularity with time. In order to enhance the reliability of a Web system, the Web administrator have to determine if there exists any relationship or correlation among different Web workload characteristics and the errors having an impact on the reliability of the Web system, so that he will be able to predict them accurately. It may not be possible to establish a generalized relationship among different Web workload characteristics. Hence, in this paper, we have performed principal component analysis (PCA) to check whether different Web workload characteristics, for particular Web software are correlated or not. Then, we have proposed a transfer function based model for Web software fault prediction. Also, we have used the pre-whitening technique to eliminate the noise present in the data for developing an efficient transfer function based model to predict the cumulative occurrences of different Web failures having an impact on the reliability of the Web software.