Abstract:
Precisely anticipating a player's position is critical to the success of team tactics and
talent scouting in football, as every position requires a unique set of talents. This
thesis uses an extensive dataset of 100,995 players and 14 important features to
explore how machine learning could be able to simplify this challenging endeavor.
Using nine different machine learning models (e.g., Random Forest, XGBoost, and
LightGBM), the research carefully trains and assesses each model's prediction
power. Under the direction of an exacting assessment methodology that includes
accuracy, precision, recall, F1 Score, and AUC-ROC Curve, the study carefully
adjusts hyperparameters to reach peak performance. With an astounding maximum
accuracy of 90.42%, the study demonstrates the great potential of machine learning
in football statistics. This research holds the potential to transform player assessment
and tactical decision-making by revealing crucial insights into the interaction
between players' locations and qualities. The ramifications go beyond the playing
field; they provide a model for data-driven insights in a number of fields where it is
essential to comprehend individual responsibilities within complex systems.
