An Adaptive Neuro-Fuzzy Model to Predict the Reliability before Testing Phase

Abstract

Reliability of a system has a direct impact on the success of any software system. Predicting the reliability at an early stage helps to optimize the testing and maintenance of the software. Rapid changes in hardware and software technologies lead to inventions of new methodologies and needs developing and validating a reliability predicting model for each method, Machine Learning approach can provide a solution to this problem. Selecting suitable metrics that affect the reliability of the system input to a model, and output of the model should reflect whether a system is reliable or not. In this paper, software is developed based on the design of Service-Oriented Architecture (SOA) methodology, is considered where a set of components interact for autonomous services. Further, to predict the reliability of a component, the probability that a component executes without fault, Adaptive Neuro-Fuzzy Inference System (ANFIS) approach is used. The coefficient of determination is evaluated to validate the model.