An Information Gain and Hierarchical Agglomerative Clustering Analysis in Identifying Key Performance Parameters in Elite Beach Soccer

The current investigation is aimed at identifying the relevant essential performance parameters (PP) that could discriminate winning and losing performance through the application of multivariate and machine learning analysis. A set of 20 different PP was collected from the Asian beach soccer tournament that constituted information such as, tactical and technical strategies, winning and losing performances. An information gain (IG) analysis is applied to extract the features that could best describe winning and losing performance. Hierarchical agglomerative cluster analysis (HACA) is used to create two different clusters based on the initial 20 PP, and the extracted features from IG whilst a canonical discrimination function analysis was used to ascertain the level of separation ability between the two aforesaid clusters. The IG identified a set of 11 PP that could best describe the winning and losing performances and the HACA formed two distinctive clusters. It is shown that the clusters formed using the 11 PP identified was able to offer an excellent separation of the two teams as opposed to the use of the initial 20 PP. The Canonical Correlation function provided a discrimination power of 0.80 for the 11 PP in comparison to the 0.70 obtained when using the 20 PP. The techniques employed in the present study serve useful in identifying key PP that could best describe winning and losing performances in the elite Asian beach soccer tournament which could assist the coaches in modifying playing strategies to ensure victory.