Prediction of Perceived Exertion Ratings in National Level Soccer Players Using Wearable Sensor Data and Machine Learning Techniques

Table 4. Results of all machine learning and deep learning models on both of our datasets of the best-performing model during the hyperparameter study. For all models, we report MAE, MSE, and RMSE of the respective 5-fold cross-validation. Best values are highlighted in bold.

Model	Matches included			Matches excluded
Model	Mean Absolute Error	Mean Squared Error	Root Mean Squared Error	Mean Absolute Error	Mean Squared Error	Root Mean Squared Error
KNeighbors	1.29 (�0.02)	3.81 (�0.09)	1.95 (�0.02)	1.21 (�0.01)	3.5 (�0.04)	1.87 (�0.01)
Ridge	1.29 (�0.02)	3.81 (�0.09)	1.95 (�0.02)	1.21 (�0.01)	3.5 (�0.04)	1.87 (�0.01)
Lasso	2.5 (�0.01)	8.25 (�0.07)	2.87 (�0.01)	2.5 (�0.03)	8.25 (�0.15)	2.87 (�0.03)
ElasticNet	2.5 (�0.01)	8.25 (�0.07)	2.87 (V0.01)	2.5 (�0.03)	8.25 (�0.15)	2.87 (�0.03)
Decision Tree	1.7 (�0.04)	7.71 (�0.28)	2.78 (�0.05)	1.54 (�0.04)	6.85 (�0.25)	2.62 (�0.05)
ExtraTree	1.25 (�0.01)	3.05 (�0.05)	1.75 (�0.01)	1.15 (�0.03)	2.65 (�0.12)	1.63 (�0.04)
Random Forest	1.47 (�0.01)	3.52 (V0.06)	1.88 (�0.02)	1.37 (�0.02)	3.12 (�0.11)	1.77 (�0.03)
Gradient Boosting	2.36 (�0.01)	7.58 (�0.08)	2.75 (�0.02)	2.35 (�0.02)	7.54 (�0.12)	2.74 (�0.02)
AdaBoost	2.4 (�0.02)	7.65 (�0.08)	2.77 (�0.01)	2.39 (�0.03)	7.54 (�0.13)	2.75 (�0.02)
DL	1.21 (�0.01)	3.47 (�0.07)	1.86 (�0.01)	1.04 (�0.07)	2.70 (�0.23)	1.64 (�0.05)