Table 4.

The RMSE of corn yield prediction performance of models.

Test date		Models
Year	Month	Ridge	Lasso	RF	DFNN	RT	3D-CNN	YieldNet
2016	July	23.12	21.03	22.48	22.16	29.41	18.84	18.73
	August	23.16	19.68	20.95	20.48	29.16	15.25	15.76
	September	24.53	20.6	21.23	21.04	29.31	16.55	15.96
	October	24.93	21.05	21.15	20.74	27.96	16.65	15.85
2017	July	30.55	27.53	26.61	26.40	33.64	22.50	20.88
	August	25.16	22.27	22.25	20.85	28.02	16.60	17.74
	September	24.15	21.5	21.99	19.21	26.8	15.71	15.53
	October	25.73	20.94	22.14	18.90	26.78	15.69	15.40
2018	July	27.51	21.21	22.38	22.85	27.69	20.64	22.08
	August	24.5	19.46	21.52	21.14	29.34	18.81	18.25
	September	25.1	18.69	21.7	20.57	28.91	17.58	16.89
	October	32.5	19.2	22.28	21.63	28.9	17.72	16.75
Average		25.91	21.10	22.22	21.33	28.83	17.71	17.49

The average ± standard deviation for corn yield in years 2016, 2017, and 2018 are, respectively, $165.72\pm 30.35$, $168.50\pm 32.88$, and $170.77\pm 34.95$. The numbers of test samples in years 2016, 2017, 2018 for corn yield prediction are 885, 882, and 784, respectively. The unit of RMSE is bushels per acre.

Best results are given in bold.