Football Player Analysis for Identifying Best Team using Machine Learning

Abstract

In the game of football (soccer), the evaluation of players for transfer, scouting, squad formation and strategic planning is important. However, due to the vast pool of grassroots level player, short career span, differing performance throughout the individual’s career, differing play conditions, positions and varying club budgets, it becomes difficult to identify the individual player's performance value altogether. The Player Performance Prediction system aims at solving this complex problem analytically and involves learning from various attributes and skills of a football player. It considers the skill set values of the football player and predicts the performance value, which depicts the scope of improvement and the capability of the player. The objective of this project is to help the coaches and team management at the grassroots as well as higher levels to identify the future prospects in the game of football without being biased to subjective conditions like club budget, competitiveness in the league, and importance of the player in the team or region. The system is based on a data-driven approach and we train our models to generate an appropriate holistic relationship between the players’ attributes values, market value and performance value to be predicted. These values are dependent on the position that the football player plays in and the skills they possess. In This project best player is predicted by algorithms namely Naïve Bayes (NB) as proposed and K Nearest Neighbor (KNN) as existing system and compared in terms of Accuracy. From the results obtained its proved that proposed NB works better than existing KNN..