Does encoding matter? A novel view on the quantitative genetic trait prediction problem

BMC Bioinformatics

Table 4 The average correlation of the top-100 markers with the highest absolute correlation values to the trait under different encoding methods. We also show the percentage of the positively correlated features for the traditional encoding

Data set	Ordinal encoding	Hybrid one	Hybrid two	Target-based
	(positive percentage)			encoding
Maize: Flint 1 TASS	0.4	0.41	0.41	0.43
Maize: Flint 2 DMC	0.35	0.37	0.37	0.38
Maize: Flint 3 DM_Yield	0.22	0.23	0.23	0.30
Maize: Dent 1 Tass	0.36	0.37	0.37	0.38
Maize: Dent 2 DMC	0.4	0.40	0.40	0.40
Maize: Dent 3 DM_Yield	0.32	0.33	0.33	0.34

ISSN: 1471-2105