Improved genomic prediction using machine learning with Variational Bayesian sparsity

Table 2 For each of the data sets, the Pearson correlation between the observed and predicted grain yield from each of the genomic prediction methods conducted using four random cross-validation splits with 90% training data and 10% validation data. The number of feature selected markers for VBS-ML is given in parentheses

Year	Methods	1	2	3	4	Ave
2014	LMM	0.52	0.49	0.50	0.47	0.50
	BayesA	0.54	0.50	0.51	0.48	0.51
	BayesB	0.55	0.51	0.51	0.49	0.51
	Naive-ML	0.54	0.41	0.39	0.42	0.44
	VBS-ML (354)	0.66	0.50	0.47	0.52	0.54
2016	LMM	0.63	0.56	0.65	0.57	0.60
	BayesA	0.63	0.56	0.65	0.58	0.61
	BayesB	0.64	0.55	0.65	0.57	0.60
	Naive-ML	0.33	0.24	0.41	0.33	0.33
	VBS-ML (409)	0.62	0.53	0.68	0.56	0.60
2017	LMM	0.48	0.52	0.53	0.52	0.51
	BayesA	0.48	0.52	0.51	0.51	0.51
	BayesB	0.48	0.51	0.51	0.53	0.51
	Naive -ML	0.33	0.38	0.40	0.52	0.41
	VBS-ML (315)	0.49	0.55	0.54	0.60	0.54
2018	LMM	0.54	0.54	0.46	0.48	0.51
	BayesA	0.54	0.54	0.47	0.47	0.50
	BayesB	0.54	0.54	0.49	0.46	0.51
	Naive-ML	0.41	0.25	0.32	0.37	0.34
	VBS-ML (385)	0.52	0.50	0.57	0.44	0.51

ISSN: 1746-4811