Table 1 Forecasting results (18-month window, aggregate by mean daily variables).

	ACT	NSW	NT	QLD	SA	TAS	VIC	WA
Number of doctor visits (Service)
AR	593.11	9,229.43	326.51	11,398.27	4,895.99	2,360.56	5,834.34	4,611.95
VAR	492.81	9,430.43	290.99	14,084.81	5,162.97	2,517.96	10,442.86	5,005.31
ML	332.62	9,799.43	165.67	6,387.70	4,634.39	1,806.72	4,770.66	3,578.60
ML-T	468.24	6,480.01	168.85	7,835.16	4,755.29	1,803.39	4,698.24	3,281.31
ML-S	291.83	4,693.23	171.84	8,162.75	4,086.66	1,104.19	3,097.64	2,701.76
ML-TS	307.48	7,961.35	173.65	8,404.30	3,646.78	1,546.50	4,795.78	2,472.54
Value of services (Benefit)
AR	8,008.59	196,404.63	3,744.19	159,764.72	67,612.19	30,815.80	151,046.58	77,726.92
VAR	8,166.70	186,727.08	3,499.06	192,864.08	86,997.05	39,057.18	191,527.00	87,439.68
ML	5,820.55	184,044.53	1,841.05	97,618.53	70,041.47	26,789.25	128,180.32	55,636.71
ML-T	5,657.46	110,829.72	1,866.14	65,216.95	50,126.84	22,124.29	107,665.09	50,138.69
ML-S	4,096.95	87,335.84	1,835.96	87,121.53	38,728.96	10,729.37	30,371.19	34,695.31
ML-T + S	4,889.68	117,465.46	1,734.43	67,398.23	42,465.21	12,896.19	32,418.57	34,684.30

Note: Root Mean Square Errors (RMSE) from each forecasting models and for each Australian States. Models in comparison are : autoregressive models (AR) of health demand, vector autoregressive Models (VAR) with health demand, google search sentiment and lockdown indicator, the XGBoost Machine Learning models (ML) with health demand, google search sentiment and lockdown indicator, the ML model with daily Google search data (ML-T), the ML model with a panel of monthly health demand from other states (ML-S), and the ML model with both daily Google search data and the panel of monthly health demand from other states (ML-TS).

Quick links

Search