A Hybrid Approach to Enhance Software Quality by Sentiment Analysis of Developer

Sentiment analysis evaluates the software developer’s emotions during product development. Quality of the real time project can be assured by the detecting the amount of code smell present in the software code snippet. In our work we have collected data from Oryx, Antlr4, Eclipse, Junit project and reduced the dimensionality using PCA and KPCA techniques. Our methodology focuses on the use of Ensemble approach for emotion analysis which is done on weekly basis. It is observed that Random Forest classifier consists of multiple numbers of decision trees which reduces the problem of over fitting and provides highest accuracy of 95%. JUnit open source project provides highest performance as the amount of clean code is more than dirty code that occurs due to positive sentiment of the developer. In our work we have assured the quality of the real time software projects by studying positive and negative emotions from which we can evaluate the type of code smell and in the further steps manual refactoring is performed by restructuring the code and quality is improved.