Study for predicting commit conflict complexity using Machine-Learning
- results-all.arff: Weka results for all the datasets, with all the ML algorithms.
- results-ann.arff: Weka results for different configurations of the Artificial Neural Network
- experiment-10fold-10rep.exp: Weka experiment configuration for all datasets and all ML algorithms. It shows the different configurations for Naïve Bayes, Artificial Neural Network and J48
- protocol.txt: Details of the labeling protocol
- jenkins: Contains the datasets for Jenkins
- voldemort: Contains the datasets for Voldemort
The datasets have been named like this:
project - type - set
- project : can be either "jenk" or "vold". Self-explanatory
- type : refers to the type of conflict considered, it can be "commit" (for merge conflict), "test" (for test failures) or "build" (for build failures)
- set : it determines if the dataset contains all features ("all") or only the ones chosen by feature selection ("filtered")