Table 9 Percentage improvement of proposed models over the benchmark for Clinic 2 across all years
From: Forecasting patient flows with pandemic induced concept drift using explainable machine learning
| Â | 2017 | 2018 | 2019 | 2020 | 2021 | 2022 | Overall |
|---|---|---|---|---|---|---|---|
Prophet | 25.8 | 12.0 | 26.2 | −25.0 | 29.6 | −15.1 | 11.2 |
Gradient Boosting | 21.6 | 12.2 | 28.7 | −0.7 | 38.0 | 17.2 | 19.7 |
Random Forest | 23.8 | 13.0 | 27.3 | 2.0 | 36.1 | 19.1 | 20.3 |
CatBoost | 23.0 | 12.0 | 26.9 | 5.1 | 38.1 | 14.0 | 20.4 |
Stacking | 27.5 | 17.0 | 33.7 | −2.6 | 37.9 | 13.6 | 21.9 |
Averaging | 29.0 | 16.8 | 34.0 | 1.1 | 39.4 | 16.1 | 23.4 |
Voting | 29.8 | 16.6 | 34.6 | −1.2 | 40.3 | 17.2 | 23.4 |