Defect prediction is a set of techniques used to identify a likely buggy software change (eg. a commit). Various measurements from previous changes are taken into consideration to predict weather a new change is likely to contain a bug or not. Commit messages or bug tracking system entries are usually examined to gather the measurements. Machine learning is often used to classify the buggy/clean changes. We are working now on adding a continuous notion to defect prediction. On one side by building on top the idea of continuous defect prediction in the IDE (Integrated Development Environment). On the other side by perfecting the prediction by using the unambiguous results of continuous integration builds of the software project.
The technique was described in a paper I coauthored under the title “Continuous Defect Prediction: The Idea and a Related Dataset”. I will present the paper at the 14th International Conference on Mining Software Repositories 2017 (MSR) in Buenos Aires, Argentina in May this year. MSR is colocated with 39th International Conference on Software Engineering (ICSE) which I will attend.