Research On Team Sports Performance Based On Network Structure

A comprehensive analysis of team sports requires an interdisciplinary approach,complemented by player-level and team-level analysis.Due to its interdisciplinary and complex characteristics,complex network theory involves the knowledge and theoretical basis of many disciplines such as system science and mathematics,which can provide a necessary theoretical framework for the analysis of team sports.Abstract team sports as a network system,one can explore the performance of individuals in the team,the type of team interaction,the interdependence between and within the team,and how to destroy the balance of the opponent team or create scoring opportunities.It can help develop strategic plans that address the tactical and technical needs of the team,thus promoting the development of team sports.This master’s thesis mainly studies the evaluation of individuals in team sports from the level of network structure,and analyze the performance of team sports and identify the cooperation mode in the team.The specific work contents are as follows:This master’s thesis summarizes and improves the common way of establishing network model for team sports.Two kinds of structural entropy are used to explore the passing network at the micro and macro levels respectively.The basketball game is constructed as a passing network containing with outcome nodes.The node entropy centrality is defined based on the node local area network.Taken player’s efficiency value as reference,three evaluation methods are selected to evaluate the index compared with several network centrality indexes.In addition,a 5-step 40-pass network sequence is constructed for each game,and focusing on the timing of the game,the correlation between network structure entropy and team performance is explored.In order to identify cooperation patterns in team sports,a motif-based spectral clustering method is proposed.In this method,the weights of weighted networks are processed by mapping,and then a motif-based spectral clustering method is extended to directed weighted networks,and the effectiveness of the clustering method is verified by experiments on a reference network and real networks.This clustering method is applied to the network formed by basketball and the results are analyzed.