A Trial of Analysis on Soccer Formations by using Self-Organizing Maps

Abstract

Time-series prediction using self-organizing maps is applied to RoboCup formation prediction. The formation of players are converted into a numerical vector, which is used as an input to the self-organizing map. The units in the self-organizing map are arranged in a two-dimensional or three-dimensional way, which allows time-series processing. A series of numerical experiments are conducted to investigate the effectiveness of the self-organizing maps.