Evaluation of maintainability with machine learning
In a research collaboration with the Technical University of Munich, we are investigating the possibilities offered by the use of machine learning for the quality assessment of software. As manual reviews take a lot of time and can therefore often only be carried out on a random basis, AI offers the opportunity to reduce effort and improve quality.
This study investigated how metrics from multiple static code analysis tools can be combined with machine learning to automatically and reliably assess the maintainability of software.
- Tools for static code analysis are widespread
- Metrics must be interpreted by experts
- Machine learning reduces the effort considerably
- Analyzed code base of 115,000 LoC
- 80% precision achieved
Read more?
LEARNING A CLASSIFIER FOR PREDICTION OF MAINTAINABILITY BASED ON STATIC ANALYSIS TOOL
Markus Schnappinger, Mohd Hafeez Osman, Alexander Pretschner, Arnaud Fietzke<br>
2019 IEEE/ACM 27TH INTERNATIONAL CONFERENCE ON PROGRAM COMPREHENSION (ICPC)
