Software quality can be assured by identifying the amount of code smell present in the open-source project. The aim of our work to improve the quality of various real-time projects which can be done by detecting the percentage of bad smell present in the internal structure of code. Dirty code analysis is done by observing the overall sentiment of software developer. In our work, we have considered MCT and Titan project developer’s emotion score that is positive, neutral, and negative sentiment values are used as the measuring parameters. By using principal component analysis (PCA) feature extractor, we have extracted the featured data which is further processed by using machine learning regression. In order to evaluate the efficiency of the project, six different regressions like Linear, Logistic, Ridge, Lasso, Polynomial, and Stepwise regression techniques are used. It is observed from our work that Lasso regression gives highest accuracy of 93%.