Code-smells identification by using PSO approach

Abstract

Code-smell defines the smells which arise in coding part of the software development life cycle. It is very crucial to detect smells in software projects. If smells are detected earlier then the possibility of occurrence of errors, faults will be reduced. Hence, quality of the software is improved. The existing work used Bayesian approaches, manual approaches and search-based approaches to detect smells. These approaches lack in getting optimization solutions in detecting process. So, paper makes use of one of the popular optimization technique called Particle Swarm Optimization (PSO) for detecting the smells in programming part. The technique shows how intelligently the smells are detected and mainly concentrated on five types of smells namely Long Methods, Long Parameters, Large Classes, Duplicated Codes, and Primitive Obsessions. Implementation of this technique is, considering source-code of any software applications or programs and injecting PSO technique into the system. Here, PSO has trained to detect five types of smells whenever their appear in the source-code. Detecting the smells in initial stages of the project gives best performance of the software, and in other hand quality of the software is achieved. Experimental results are shown by using PSO technique, where searching time will be less consumed and accuracy of the system is gained.