Custom machine learning classifiers for study selection in systematic reviews or maps of research: a living study within a review (SWAR)

This project is a living study within a review (living SWAR) that aims to assess the performance of custom binary machine learning classifiers for selecting eligible records to be retained for screening for potential inclusion in systematic reviews or maps of research. This .osf project page (https://osf.io/f85m7/) hosts a repository of datasets from each replication ('implementation') of this SWAR (see 'Files' tab). It will also host datasets from the prospective meta-analysis components of this living SWAR when ready. Datasets from three replications of this SWAR completed to date are currently available in this repository (see 'Files' tab): Implementations 001, 002 and 003. Datasets from prospective meta-analysis components have not yet been added. Templates for reporting the results of each replication of this SWAR are also available in the 'Files' tab. This repository was last updated with new replications on 7th January 2025. A copy of the SWAR registration form / protocol for this living SWAR is also published in the 'Files' tab (accepted version - December 2025). This SWAR was pre-registered in the SWAR Repository Store in December 2025: https://www.qub.ac.uk/sites/TheNorthernIrelandNetworkforTrialsMethodologyResearch/SWATSWARInformation/Repositories/SWARStore/ For all enquiries, contact: Ian Shemilt (i.shemilt@ucl.ac.uk).