Table 5 Accuracies and mean ranks for all models across each year for Clinic 1
From: Forecasting patient flows with pandemic induced concept drift using explainable machine learning
| Â | 2017 | 2018 | 2019 | 2020 | 2021 | 2022 | Mean R | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Â | MAPE | R | MAPE | R | MAPE | R | MAPE | R | MAPE | R | MAPE | R | |
kNN | 16.8 | 13.2 | 16.1 | 13.3 | 15.4 | 13.4 | 22.7 | 11.8 | 16.1 | 12.2 | 17.8 | 12.3 | 12.7 |
SVR (NU) | 17.2 | 13.8 | 15.5 | 12.9 | 14.9 | 13.2 | 27.2 | 11.6 | 15.2 | 10.9 | 15.2 | 11.5 | 12.4 |
Benchmark | 12.6 | 10.3 | 12.9 | 11.3 | 11.7 | 11.8 | 29.9 | 12.4 | 17.5 | 11.4 | 14.8 | 10.8 | 11.4 |
Naive | 13.4 | 11.1 | 12.6 | 10.9 | 12.1 | 11.5 | 25.6 | 11.0 | 18.7 | 11.1 | 14.1 | 9.9 | 11.0 |
Naive (Enhanced) | 10.0 | 8.2 | 9.3 | 7.5 | 8.6 | 7.3 | 14.9 | 8.0 | 11.1 | 7.7 | 9.7 | 6.3 | 7.6 |
ARIMA | 9.7 | 7.0 | 9.7 | 8.7 | 9.0 | 7.7 | 13.1 | 7.2 | 10.8 | 6.7 | 10.5 | 6.8 | 7.4 |
Random Forest | 9.3 | 7.7 | 9.0 | 7.0 | 8.1 | 6.9 | 13.9 | 7.3 | 9.9 | 7.0 | 9.8 | 6.0 | 7.1 |
Kernel Ridge | 9.0 | 6.2 | 9.2 | 7.1 | 7.9 | 5.9 | 12.5 | 6.9 | 10.6 | 8.2 | 10.7 | 8.8 | 7.0 |
Ridge | 9.0 | 6.2 | 9.2 | 7.2 | 7.9 | 6.0 | 12.4 | 6.6 | 10.4 | 7.7 | 10.8 | 8.9 | 6.9 |
Prophet | 8.4 | 4.9 | 8.4 | 5.8 | 8.2 | 7.1 | 13.5 | 8.1 | 10.2 | 7.3 | 14.0 | 9.8 | 6.9 |
CatBoost | 9.7 | 8.7 | 8.7 | 6.3 | 8.1 | 7.0 | 11.9 | 5.8 | 9.7 | 6.8 | 9.7 | 6.0 | 6.8 |
Gradient Boosting | 9.2 | 7.5 | 8.9 | 6.5 | 8.0 | 6.6 | 12.2 | 6.1 | 9.8 | 6.5 | 9.7 | 6.3 | 6.6 |
Stacking | 8.4 | 4.9 | 8.3 | 5.2 | 7.8 | 5.8 | 12.2 | 6.5 | 9.7 | 6.6 | 11.1 | 6.4 | 5.8 |
Averaging | 8.5 | 5.5 | 8.2 | 4.8 | 7.6 | 5.0 | 11.9 | 5.4 | 9.2 | 5.0 | 9.9 | 5.3 | 5.1 |
Voting | 8.4 | 4.8 | 8.2 | 5.4 | 7.5 | 4.8 | 11.6 | 5.3 | 8.9 | 4.8 | 9.7 | 4.8 | 5.0 |