Unleashing potential: a data-driven exploration of identifying player potentialities through advanced analytics in sports
Abstract
In this project, we delve deeper into the complex area of predicting a potential replacement
of a footballer of a specific position. For that, we used multiple machine
learning models on the Sofifa dataset. Our analysis reveals interesting insights into
the predictive capabilities of these models, with a particular focus on numerical performance
measures. Among the models tested, the LightGBM Regressor appears
to be the epitome of predictive power. This algorithm consistently outperforms the
others, showing the lowest mean squared error (MSE) and highest R-squared value
on both the test data set and the overall data set. Her ability to navigate the complexity
of player performance patterns is evident, making her a leader in our prediction
arsenal. Complements for Random Forest Regressor, XGBRegressor Regressor,
LightGBM Regressor, and CatBoost Regressor demonstrate superior performance,
characterized by consistently low MSE values and high R-squared values. These
gradient boosting algorithms demonstrate their effectiveness in capturing complex
patterns in the Sofia dataset. The Linear Regressor model utilizes its power in understanding
the linear releationships among the data and gives a higher accuracy
too. The KNeighbors-Regressor, with its proximity-based approach, also achieves
stripes, especially by achieving high R-squared values. This model excels at identifying
players with similar characteristics, highlighting their collective impact on
overall performance. It should be noted that, Support Vector Regressor (SVR) and
Neural Network models provide valuable insights, despite relatively lower prediction
accuracy. These models highlight the complexity inherent in player forecasting and
highlight the need for meticulous parameter tuning. LightBGM Regressor stands
out as the superior model for predicting our research, closely followed by XGBRegressor,
Random Forest Regressor, CatBoost Regressor. These results highlight the
importance of selecting models that match the variation of the data set to accurately
and reliably predict performance in soccer analytics.