Table 4: Accuracy of the models of patient arrivals at ED2 for different forecasting intervals.
| Interval |
Model |
MAE |
MAPE |
MASE |
| In |
Out |
In |
Out |
In |
Out |
| 2 hours |
|
|
|
|
|
|
|
|
Regression |
2.75 |
2.54 |
46 |
37 |
0.77 |
0.71 |
|
ARIMA (2,0,2)×(1,0,1)84
|
2.65 |
2.37 |
43 |
41 |
0.74 |
0.66 |
|
Naïve model |
3.56 |
3.30 |
54 |
50 |
1.00 |
0.93 |
| 4 hours |
|
|
|
|
|
|
|
|
Regression |
4.19 |
3.93 |
41 |
34 |
0.81 |
0.76 |
|
ARIMA (3,0,3)×(0,1,1)42
|
3.79 |
3.36 |
34 |
34 |
0.73 |
0.65 |
|
Naïve model |
5.19 |
4.60 |
43 |
39 |
1.00 |
0.89 |
| 8 hours |
|
|
|
|
|
|
|
|
Regression |
6.62 |
6.57 |
35 |
26 |
0.85 |
0.84 |
|
ARIMA (1,0,4)×(1,1,1)21
|
5.77 |
5.43 |
28 |
21 |
0.74 |
0.70 |
|
Naïve model |
7.80 |
7.49 |
34 |
30 |
1.00 |
0.96 |
| 24 hours |
|
|
|
|
|
|
|
|
Regression |
10.97 |
11.18 |
12 |
9.9 |
0.72 |
0.74 |
|
ARIMA (0,1,2)×(1,0,1)7
|
10.86 |
13.02 |
12 |
11 |
0.72 |
0.86 |
|
Naïve model |
15.16 |
12.03 |
16 |
11 |
1.00 |
0.79 |
Note: MAE = Mean Absolute Error, MAPE = Mean Absolute Percentage Error, MASE = Mean Absolute Scaled Error, In = in sample (i.e., calculated over the data used in building the model: January 2012 – December 2014), Out = out of sample (i.e., calculated over the forecasted data: January 2015).