Predicting the Number of Reported Pulmonary Tuberculosis in Guiyang, China, Based on Time Series Analysis Techniques

Table 5

Verification of alternative ARIMA models.

ARIMA models	Training set			Test set			AIC	Box-Ljung test
	MAPE (%)	RMSE	MAE	MAPE (%)	RMSE	MAE		-squared	value
ARIMA (1, 0, 1) (0, 1, 1)12	11.102	39.983	31.659	12.436	32.012	29.046	1131.050	0.301	0.584
ARIMA (0, 1, 1) (0, 1, 0)12	12.726	48.733	36.806	13.187	44.787	32.290	1152.190	0.101	0.750
ARIMA (1, 1, 1) (0, 1, 0)12	12.712	48.660	36.901	13.665	46.234	33.447	1153.830	0.005	0.942
ARIMA (0, 1, 1) (0, 1, 1)12	10.549	39.956	30.731	7.733	26.653	20.055	1122.370	0.105	0.746
ARIMA (1, 1, 0) (0, 1, 1)12	11.012	42.498	32.325	7.226	28.546	19.206	1132.510	0.779	0.378
ARIMA (1, 1, 1) (0, 1, 1)12	10.462	39.535	30.411	9.936	27.894	24.205	1123.800	0.092	0.762
ARIMA (1, 1, 1) (1, 1, 1)12	10.319	38.826	29.973	8.034	26.612	20.406	1124.940	0.048	0.826
ARIMA (0, 1, 1) (0, 1, 2)12	10.271	38.819	29.906	7.229	27.913	19.165	1122.680	0.044	0.833
ARIMA (0, 1, 1) (1, 1, 2)12	10.307	39.062	30.015	7.179	27.852	19.075	1124.440	0.055	0.814
ARIMA (0, 1, 1) (2, 1, 2)12	8.399	32.532	24.309	14.232	41.408	34.847	1115.750	0.001	0.979
ARIMA (1, 1, 1) (0, 1, 2)12	10.184	38.375	29.578	7.081	26.771	18.601	1124.440	0.030	0.863
ARIMA (1, 1, 1) (1, 1, 2)12	10.277	38.692	29.817	7.302	26.625	19.071	1126.150	0.036	0.849

Note: ARIMA (1, 0, 1) (0, 1, 1)₁₂ is the autorecognition model of auto.arima() function.