Design Smell Detection and Analysis for Open Source Java Software

Abstract

Software design smells have gained significant importance in recent years since those directly lead to the increase of design debts and drastically affect software quality. Although the impact of design smells is manifold, techniques to detect design smells using both rule based and data mining approaches have been explored to a limited extent. This research aims to provide a tool which uses software metrics as a guide to detect smells and also deploys Spectral Clustering to mine the software repositories and group similar smells. The tool has been partially implemented till now and initial experiments on 2,59,509 Lines of Code (LoC) covering 3,306 classes of real life open source Java software show 2,220 occurrences of four types of design smells.