축구 월드컵대회의 경기기록 기반 경기결과 예측

Abstract

This study was to confirm the probability of prediction of match outcomes based on the official data of soccer World Cup. Also, it was to identify the differences of prediction models which were divided with the ratio of data used among the raw data. The raw data was considered in the official data from 2002 to 2018 soccer World Cup on the official web sites of FIFA. Totally, 17 independent variables were shot attempts, shot on goal, % of shot on goal, passes, passes completed, % of passes completed, short passes, short passes completed, % of short passes completed, long passes, long passes completed, % of long passes completed, corner kicks, fouls committed, fouls suffered, offsides, and % of ball possessions. One dependent variable was the outcome of match. In order to process the data, the self-organizing map which is one of artificial intelligent techniques was used for this study. Totally, 640 data set was used for this study that 60%, 70%, 80%, and 90% of raw data was split as the training data set for the self-organizing map. The results of this study were found as following belows; First, there were a couple of consideration on designing of the self-organizing map that the error was 3.446 and the structure was 10 × 10. Second, 60% (384) and 90% (576) of usage on the raw data for the training matched the prediction with 71.86%. Third, 70%(448) and 80%(512) of usage on the raw data for the training shown higher prediction with 73.44%. Consequently, the results of this study were shown that there were no differences of prediction accuracy with different amount of data used.