BugBang

This repository contains an R script to classify an issue report from JIRA as referring to a bug or not. It also contains associated datasets. They are used in the following paper:

Nitish Pandey, Debarshi Kumar Sanyal, Abir Hudait, and Amitava Sen, “Automated Classification of Software Issue Reports Using Machine Learning Techniques: An Empirical Study,” Innovations in Systems and Software Engineering, Springer. (Accepted, March 2017) (doi:10.1007/s1133)

Please cite the above paper if you use the scripts in this repository. I thank Prof. Hideaki Hata for the data (which were then augmented with records from JIRA).

Contents

BugBang_rel_v4.0.R => Main R script for issue classification

datasets/exp1 => Datasets for experiment 1

datasets/exp2 => Datasets for experiment 2

LogsForJournal.zip => Logs containing results reported in the journal paper

TablesAndGraphsForJournal.zip => Tables, graphs and scripts (to plot graphs from the tables) for results reported in the journal

Commandline in Linux

Rscript BugBang_rel_v4.0.R --infile="./datasets/exp1/http_client.csv" --outfile="./out/classification_http_client.out" --max_terms_in_dtm=0.25 --normalize=3 --cv_fold=10

Ensure that the input files and the output directory exist.

Script Modification for Experiment 2

The released script is for experiment 1 (data in folder "exp1"). For experiments in exp2, "CLASSIFIED" in the following lines should be replaced with "TYPE":

training.data.input <- training.data.input [ training.data.input$CLASSIFIED %in% c("BUG", "NUG"), ]

Train.Type <- as.factor(training.data.input$CLASSIFIED);