Identification of Methylation Signatures and Rules for Sarcoma Subtypes by Machine Learning Methods

Table 1

Performance of random forest and decision tree under some feature subsets derived from three lists generated by three feature ranking algorithms.

Feature ranking algorithms	Classification algorithm	Number of features	Weighted	Macro	MCC	ACC
LASSO	RF	110	0.971	0.980	0.971	0.971
	RF	8850	0.987	0.990	0.987	0.987
	DT	7490	0.867	0.863	0.863	0.867
LightGBM	RF	30	0.940	0.948	0.939	0.940
	RF	4520	0.991	0.993	0.990	0.990
	DT	3900	0.868	0.861	0.864	0.868
MCFS	RF	310	0.978	0.982	0.978	0.978
	RF	7020	0.987	0.991	0.987	0.987
	DT	5520	0.868	0.860	0.863	0.866