Developing a Ranking Problem Library from a data-oriented perspective

We present an improved library for the ranking problem called RPLIB. RPLIB includes the following data and features. (1) Real and artificial datasets of both pairwise data (i.e., information about the ranking of pairs of items) and feature data (i.e., a vector of features about each item to be ranked). These datasets range in size (e.g., from small $n=10$ item datasets to large datasets with hundred of items), application (e.g., from sports to economic data), and source (e.g. real versus artificially generated to have particular structures). (2) RPLIB contains code for the most common ranking algorithms such as the linear ordering optimization method and the Massey method. (3) RPLIB also has the ability for users to contribute their own data, code, and algorithms. Each RPLIB dataset has an associated .JSON model card of additional information such as the number and set of optimal rankings, the optimal objective value, and corresponding figures.