Bayesian Inference to Predict Smelly classes Probability in Open source software

Abstract

Software testing entails a number of processes that are focused on finding faults within a stipulated time. Lots of papers have been published for object oriented metrics but mostly concentrating on software fault prediction, very few has been published for bad smells. Bad code smells are used to recognize complex classes in object-oriented software systems for refactoring. This study contributes to all code smell prediction techniques by designing a Logistic regression model and using Bayesian inference graphs. This paper shows the results of a study in which Object Oriented metrics effectively predict design smell for an open source system. Software metrics assess as predictor of smelly classes. Bayesian inference graphs can represent decision for finding the smells present in software system. For Probabilistic reliability analysis, Bayesian inference is intended to be used for risk related data. This paper presents the relationship between smelly classes and object-oriented metrics. This study demonstrates a statistical technique for estimating the smelly classes for any piece of software. We examined the open source Eclipse system, which has a strong industrial usage. Our main objective is to design a Bayesian Inference graph to predict bad smell in the code.