Hybrid Parameter Optimization Approach with Adaptive Neuro Fuzzy Inference System for the Software Maintainability

Abstract

This paper presents a novel method to measure the maintainability of the software from the design artifact. It is an inevitable measure because it aims to attain software with a better quality. The system is designed to measure the maintainability of the system from the UML class metric. This is extracted from the UML class diagram to predict the maintainability of the class diagram. The system is implemented using CFS from the Weka tool to select an optimized variable from a set of variables i.e UML class metric. Hybrid ANFIS is an artificial intelligence technique which has been incorporated with the optimizing algorithms to reduce the overall number of UML metric and build a Fuzzy Inference System (FIS) based on the learning process. The optimization attains an enhanced result since it is done continually by both using feature selection and optimization algorithms repetitively, which results in reducing the UML metric considerably to measure the maintainability of the software. The proposed research work is evaluated in terms of the performance measures, MSE, RMSE, true positive rates and the result is clearly shown that a better optimization of the maintainability measure estimation process can be done.