Hybrid neural network approach for predicting maintainability of object-oriented software

Abstract

Estimation of different parameters for object-oriented systems development such as effort, quality, and risk is of major concern in software development life cycle. Majority of the approaches available in literature for estimation are based on regression analysis and neural network techniques. Also it is observed that numerous software metrics are being used as input for estimation. In this study, object-oriented metrics have been considered to provide requisite input data to design the models for prediction of maintainability using three artificial intelligence (AI) techniques such as neural network, Neuro-Genetic (hybrid approach of neural network and genetic algorithm) and Neuro-PSO (hybrid approach of neural network and Particle Swarm Optimization). These three AI techniques are applied to predict maintainability on two case studies such as User Interface System (UIMS) and Quality Evaluation System (QUES). The performance of all three AI techniques were evaluated based on the various parameters available in literature such as mean absolute error (MAE) and mean Absolute Relative Error (MARE). Experimental results show that the hybrid technique utilizing Neuro-PSO technique achieved better result for prediction of maintainability when compared with the other two.