Thesis

5B 130 CHAPTER 5B Table 2. Performance of Random Forest and Naïve Bayes Algorithm Features Accuracy Precision Recall F1 algorithm AUC Injury prediction Precision Recall F1 Injury Random Forest All* 0.89 0.03 0.22 0.88 0.52 No injury 0.98 0.91 0.95 Injury 0.03 0.11 0.05 Bayes All 0.37 0.02 1.00 0.53 0.60 No injury 1.00 0.37 0.54 Injury 0.2 1.00 0.04 Random Forest Limited# 0.87 0.04 0.125 0.91 0.43 No injury 0.97 0.86 0.92 Injury 0.04 0.12 0.06 Naïve Bayes Limited 0.41 0.03 0.83 0.56 0.60 No injury 0.99 0.40 0.57 Injury 0.03 0.83 0.6 * All features are (i)the Acute:Chronic Workload Ratio, four weeks, three weeks, two weeks, one week before either an injury or not, (ii) the average Acute:Chronic Workload Ratio over two weeks before either an injury or not, (iii) the average Acute:Chronic Workload Ratio over two weeks, two weeks before either an injury or not. #The limited number of features are: the Acute:Chronic Workload Ratio, four weeks, three weeks, two weeks, before either an injury or not UAC = Area Under Curve

RkJQdWJsaXNoZXIy MjY0ODMw