Can Machine Learning Support the Selection of Studies for Systematic Literature Review Updates?

Artifacts for "Can Machine Learning Support the Selection of Studies for Systematic Literature Review Updates?".

File used to answer RQ1:

RQ1-RF-predictions.csv

RQ1-RQ3-best-configuration-RF.csv

File used to answer RQ2:

RQ2-SVM-predictions.csv

RQ2-best-configuration-SVM.csv

File used to answer RQ3:

RQ3-RF-normalized-predictions.csv

RQ1-RQ3-best-configuration-RF.csv

The file assessment-team-votes.csv contains the title of each study, a bolean indicating if it was included or not and the individual marks of each reviewer before applying the agreement criteria.

The .bib files used in our experiment are available at:

Our testing set: 'Testing set - Excluded.bib' (513 studies) and 'Testing set - Included.bib' (38 studies). All of the 551 studies we used, were obtained from the actual SLR Update

Our training set: 'Training set - Excluded.bib' (83 studies - obtained by performing the backward snowballing using the Original SLR) and 'Training set - Included.bib' (45 studies - all studies that were included in the Original SLR).

All of our code is available in the .zip file. Besides our pipeline, there's also some jupyter notebooks in code/analysis showing illustrating how we answered each of our questions.