Prediction and detection of code smells in software

Abstract

Software is susceptible to regular modifications to suit the new specifications, which leads to bad smells. Bad smells suggest the flaws in software design. The existence of bad smells has an adverse effect on many aspects of the software like: quality, maintenance, readability and reliability. With the excessive amount of bad smells, a software system is very difficult to manage and evolve. In available studies, researchers mainly emphasized the strategies for detecting bad smells and the explanations for the evolution of bad smells in the software systems. Furthermore, researchers analyze the data residing in the repositories of software and examine the maintenance activities which are hindered by bad smells. As per our knowledge; mathematical model for bad smell detection or prediction is not available in the existing literature. The main objective of this work is to identify the bad smells during the early phase of the software life cycle. A bad smell prediction model has been proposed using: the information or Entropies: Shannon, R´enyi and Tsallis entropy. The model is validated using goodness of fit parameters (prediction error, bias and variation) and performance statistics (R-square, adjusted R-square and standard error). The secondary goal is to determine the bad smells detection rules associated with different programming languages using software metrics through machine learning.