Predicting Source Code Quality with Static Analysis and Machine Learning

  • Vera Barstad Faculty of Engineering and Science, University of Agder
  • Morten Goodwin Faculty of Engineering and Science, University of Agder
  • Terje Gjøsæter Faculty of Engineering and Science, University of Agder

Abstract

This paper is investigating if it is possible to predict source code quality
based on static analysis and machine learning. The proposed approach
includes a plugin in Eclipse, uses a combination of peer review/human
rating, static code analysis, and classification methods. As training data,
public data and student hand-ins in programming are used. Based on
this training data, new and uninspected source code can be accurately
classified as “well written” or “badly written”. This is a step towards
feedback in an interactive environment without peer assessment.

Section
Artikler