FLAGR: A flexible high-performance library for rank aggregation

Abstract

The fusion of multiple preference lists into a single aggregate list with improved element ranking is a well-studied research area with numerous applications in bioinformatics, information retrieval, collaborative filtering, and election systems. Despite the existence of a large number of rank aggregation methods, only a small portion of them have publicly available implementations. In this paper we introduce FLAGR, a high performance, modular, open source library for rank aggregation. The library contains efficient implementations of both baseline and state-of-the-art algorithms that receive multiple ranked preference lists and output a single consensus ranking. We also introduce PyFLAGR, a library that links to the FLAGR core and allows the invocation of its implementations from standard Python programs. The package also includes a special tool that can be used to evaluate and compare the performance of the underlying algorithms. In contrast to other solutions, FLAGR has been created with flexibility in mind: third-party researchers and analysts may easily integrate their implementations into the library by developing only a single function. These features render FLAGR and PyFLAGR an attractive research platform for developing, comparing and evaluating rank aggregation algorithms. Extended descriptions of the library components and useful code examples are provided in the accompanying user manual (provided as supplemental material) and the supporting Web site at https://flagr.site.