Disciplines
Statistics and Probability
Abstract (300 words maximum)
My project improves upon the tmanual cluster method of state interleague scheduling used by the Georgia State Soccer Association by use of an automated multi-round linear assignment algorithm based on SAS Optgraph.
Currently, interleague schedules are formed by manually grouping teams according to placement on a map. Typically, ten teams are assigned to a group and each team plays all of the others once per season (nine games). This approach has two flaws: 1) distance on a map does not always correlate with travel time, and 2) assigning teams to clusters precludes teams from playing teams from other groups. The second flaw becomes particularly apparent when the ideal boundaries between clusters are not obvious, but they must be drawn based on a ten-team cluster.
The improved system uses travel time between venues as the edge weight and multiple rounds of linear-assignment mapping to generate a full schedule. As Optgraph only supports bipartite linear mapping, the nodes are randomly partitioned between an A and a B subgraph, which is shuffled after each round. This approach is also compared to a non-bipartite matching algorithm written in Python.
Academic department under which the project should be listed
CCSE - Data Science and Analytics
Primary Investigator (PI) Name
Dr. Joe DeMaio
Included in
Improved Regional Sports Scheduling using SAS Optgraph
My project improves upon the tmanual cluster method of state interleague scheduling used by the Georgia State Soccer Association by use of an automated multi-round linear assignment algorithm based on SAS Optgraph.
Currently, interleague schedules are formed by manually grouping teams according to placement on a map. Typically, ten teams are assigned to a group and each team plays all of the others once per season (nine games). This approach has two flaws: 1) distance on a map does not always correlate with travel time, and 2) assigning teams to clusters precludes teams from playing teams from other groups. The second flaw becomes particularly apparent when the ideal boundaries between clusters are not obvious, but they must be drawn based on a ten-team cluster.
The improved system uses travel time between venues as the edge weight and multiple rounds of linear-assignment mapping to generate a full schedule. As Optgraph only supports bipartite linear mapping, the nodes are randomly partitioned between an A and a B subgraph, which is shuffled after each round. This approach is also compared to a non-bipartite matching algorithm written in Python.