Table 5 Model comparison where gradient boosting was selected as the best fit
From: Understanding EMS response times: a machine learning-based analysis
Selected | Number Of Observations | Percentile | Predicted Average | ASE | Observed Average | SSE | Model |
|---|---|---|---|---|---|---|---|
Yes | 490,289 | 5 | 167.80 | 1690.09 | 172.90 | 828634299.32 | Training: Gradient boosting |
Yes | 245,144 | 5 | 168.22 | 1789.91 | 169.40 | 438785173.01 | Validation: Gradient boosting |
Yes | 81,715 | 5 | 167.29 | 1769.54 | 170.06 | 144598368.52 | Test: Gradient boosting |
No | 490,289 | 5 | 168.66 | 1307.06 | 194.87 | 640835098.42 | Training: Forest |
No | 245,144 | 5 | 166.32 | 1796.18 | 168.69 | 440322314.81 | Validation: Forest |
No | 81,715 | 5 | 165.52 | 1786.66 | 169.21 | 145997039.91 | Test: Forest |
No | 222,482 | 5 | 167.41 | 1788.88 | 169.01 | 397994152.56 | Training: Neural network |
No | 110,779 | 5 | 168.11 | 1861.60 | 166.61 | 206226675.73 | Validation: Neural network |
No | 37,213 | 5 | 168.51 | 1832.01 | 166.76 | 68174404.09 | Test: Neural network |
No | 222,482 | 5 | 129.17 | 2116.45 | 158.05 | 470871881.72 | Training: Linear regression |
No | 110,787 | 5 | 129.46 | 2138.34 | 157.61 | 236900262.60 | Validation: Linear regression |
No | 37,214 | 5 | 129.65 | 2127.13 | 157.27 | 79158981.29 | Test: Linear regression |