Semester

Fall

Date of Graduation

2020

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Lane Department of Computer Science and Electrical Engineering

Committee Chair

Hany Ammar

Committee Co-Chair

Dale Dzielski

Committee Member

Katerina Goseva-Popstojanova

Abstract

Paying-off the Architectural Technical Debt by refactoring the flawed code is important to control the debt and to keep it as low as possible. Project Managers tend to delay paying off this debt because they face difficulties in comparing the cost of the refactoring against the benefits they gain. For these managers to decide whether to refactor or to postpone, they need to estimate the cost and the efforts required to conduct these refactoring activities as well as to decide which flaws have higher priority to be refactored among others.

Our research is based on a dataset used by other researchers in the technical debt field. It includes more than 18,000 refactoring operations performed on 33 apache java projects. To estimate the refactoring efforts done, we applied the COCOMO II:2000 model to calculate the refactoring cost in person-months units per release. Furthermore, we investigated the correlation between the refactoring efforts and two static code metrics of the refactored code, mainly, the LOC and the complexity. The research revealed a moderate correlation between the refactoring efforts and each one of the size of the project and code complexity. Finally, we applied the DesigniteJava tool and machine learning practices to verify our research results. From the analysis we found a significant correlation between the ranking of the architecture smells and the ranking of refactoring efforts for each package.

Using machine learning practices, we took the architecture smells level and the code metrics of each release as an input to predict the levels of the refactoring effort of the next release. We calculated the results using our model and found that we can predict the higher refactoring cost levels with 93\% accuracy.

Embargo Reason

Publication Pending

Download

Share

COinS