Matching data sets

Here is a list of all the data sets in the matching category, showing 20 per page.

AAMAS Bidding Data

This dataset contains the bids of reviewers over papers from the 2015 and 2016 Autonomous Agents and Multiagent Systems Conference. Inclusion in these data sets were explicitly opt-in; 2015 contains 9,817 bids of 201 reviewers over 613 papers; this represents about 40% of the actual 22,360 bids of 281 reviewers over 670 papers. The 2016 data contains 161 out of 393 reviewers with bids over 442 out of 550 papers.

The bidding language for these conferences is yes/maybe/no/conflict. In order to make these more useful for PreLib users, we have converted them to incomplete partial orders of the form {yes} > {maybe} > {no response} > {no}. The papers for which a reviewer had a conflict have been removed from their preference list. All reviewers had different preference orderings, hence each file contains as many entries as reviewers. We are deeply grateful to the IFAAMAS board and Rafael Bordini, Edith Elkind, John Thangarajah, and David Shield for approving, coordinating, and providing this dataset.

Want to have more details, see this page.

Click here to download the dataset

Computer Science Conference Bidding Data

This dataset contains the bidding data from 3 Computer Science Conferences. This contains the bids of all reviewers (aside a small number of opt-outs) over a subset of papers at the conference.

The bidding language for these conferences is yes/maybe/conflict. In order to make these more useful for PreLib users, we have converted them to incomplete partial orders of the form {yes} > {maybe} > {no response}. The papers for which a reviewer had a conflict have been removed from their preference list. All reviewers had different preference orderings, hence each file contains as many entries as reviewers.

Want to have more details, see this page.

Click here to download the dataset

Kidney Data

This dataset contains 310 instances of synthetic kidney donor pools. The data was generated using a state of the art donor pool generation method (described in Saidman et al., Increasing the opportunity of live kidney donation by matching for two-and three-way exchanges. Transplantation 81(5), 2006) and was donated by John Dickerson. John has recently posted his generation as well as his exchange solving code online; it is available here.

The dataset consists of 10 randomly generated instances of kidney exchanges with 16, 32, 64, 128, 256, 512, 1024, 2048 patients and, as a percentage of the pool, altruists at 0%, 5%, 10%, and 15% for a total of 310 data files. The main components use the wmd data format described in the Data Formats page. Each edge has a source and multiple destinations to represent the patients that can receive a kidney from the source. All edges have weight 1 unless they connect from a patient to an altruist (who does not need a kidney), which have weight 0.

There is a dat file associated with each kidney exchange datafile. This file contains some extra fields that may be of interest to researchers. Specifically, the file contains the following files: Pair index number of the pair in the corresponding wmd file.; Patient the blood type of the person needing the kidney; Donor the blood type of the person donating the kidney; Wife-P? 1 if the person needing the kidney is the wife of the donor; %Pra denotes the panel reactive antibody level of the patient, discretized into three levels; Out-Deg the number of nodes in the wmd file that can receive a kidney from this donor; Altruist1 if the corresponding pair is an altruist.

Want to have more details, see this page.

Click here to download the dataset

Project Bidding Data

This dataset contains bids of students over a set of projects for student/project allocations at the School of Computing Science, University of Glasgow. Each project is supervised by an individual each with a maximum capacity of supervision. There are 8 years worth of data in this set and with between 31 and 51 students and 56 and 155 projects. This data was kindly donated by David Manlove who collected this data.

In addition to the strict and incomplete preference profiles of the students we have extended the profiles with all unranked items tied at the end. We have also posted .dat files containing the supervisor identifiers and capacities. The format for the .dat files is Supervisor ID, Capacity, Projects; where Projects is a space separated list of the projects supervised by the Supervisor. Each project has a capacity of 1 while each supervisor has a variable capacity. In academic sessions 2007-08 and 2008-09 there were no supervisor capacities in force, thus the projects and supervisors are in 1-1 correspondence.

Want to have more details, see this page.

Click here to download the dataset