Team Recommender In PES2018 Using Submodular Function Optimization

With the development of Internet and computer game technologies,recommendation systems,especially team recommendation techniques,have a wide range of applications in machine learning and data mining,and the recommendation system has become an important content in sports games.Pro Evolution Soccer 2018(PES2018)is a very popular sports game that can fully simulate real football matches.If the system can automatically compose a team with a high winning percentage,it will greatly optimize the game engine and the game experience of player.There are many factors that influence the outcome of a football match.This article believes that when the team chooses more skilled players,it is easier to win the game,this means that the strength of the team is extremely related to the coverage of each skill.Therefore,in the case of ensuring the position of the player on the field is appropriate,we proposes a football skill coverage function as an objective function to quantify the strength of the team.Then,in order to obtain the optimal solution,the submodularity of the objecti-ve function is proved,so that the football team composition problem can be modeled as a submodular function maximization problem with the total salary of the team players as the constraint.To solve the submodular function optimization problem,we propose Cost-Effective Forward selection Greedy(CEFG)algorithm,combined with generalized greedy algorithm and unit cost greedy algorithm,to improve the traditional greedy algorithm.Then we simulate football game experiment in PES2018,and demonstrate the performance of our techniques based on the statistics and analysis of the results.