Table 1 Result of two-sample T-test on the prediction accuracy on all dates of the extended and original WKDE models.

No. of days between base date and date of prediction		1	2	3	4	5	6	7	8	9	10	11	12	13	14
Accuracy of extended WKDE	Mean	0.773	0.759	0.745	0.734	0.725	0.700	0.706	0.673	0.655	0.593	0.549	0.512	0.512	0.459
Accuracy of extended WKDE	Variance	0.084	0.075	0.071	0.065	0.060	0.059	0.059	0.056	0.055	0.057	0.052	0.049	0.045	0.043
Accuracy of original WKDE	Mean	0.679	0.665	0.657	0.650	0.644	0.616	0.625	0.590	0.574	0.515	0.471	0.435	0.428	0.385
Accuracy of original WKDE	Variance	0.082	0.074	0.070	0.063	0.055	0.057	0.055	0.056	0.053	0.056	0.051	0.050	0.045	0.042
T		1.988	2.106	2.016	2.031	2.069	2.119	2.099	2.136	2.129	2.017	2.096	2.105	2.438	2.210
P(H₀)		0.049	0.037	0.046	0.044	0.040	0.036	0.038	0.034	0.035	0.046	0.038	0.037	0.016	0.029

The null hypothesis was H₀: μ(accuracy_extended) ≤ μ(accuracy_original), i.e., the extended WKDE model is not more accurate than the original WKDE model. At the significance level of 0.05, H₀ was rejected consistently for the prediction accuracy with 1–14 days between the base date and the prediction date, thus showing that the extended WKDE model has a statistically significant higher accuracy. N = 75 samples on every date were use in each two-sample T-test.

Quick links

Search