Metabolomic Biomarker Identification in Presence of Outliers and Missing Values

Table 1

Performance measurement of different methods using average RMSE between original and reconstructed data matrix for different percentage of missing values in absence and presence of outliers using simulated datasets.

Methods with different conditions		Without outliers	3% outliers	5% outliers	7% outliers	10% outliers	15% outliers
For 10% missing values	Zero	1.56	1.92	2.10	2.35	2.48	2.69
	kNN	0.77	1.42	1.71	2.05	2.23	2.43
	RF	0.82	1.45	1.73	2.07	2.24	2.45
	Proposed	0.77	1.15	1.19	1.24	1.37	1.48
For 15% missing values	Zero	2.19	2.42	2.56	2.72	2.91	3.22
	kNN	0.91	1.47	1.77	2.05	2.36	2.51
	RF	0.97	1.51	1.80	2.08	2.39	2.52
	Proposed	0.92	1.24	1.28	1.34	1.50	1.64
For 20% missing values	Zero	2.51	2.71	2.88	2.96	3.14	3.37
	kNN	1.00	1.58	1.94	2.12	2.45	2.69
	RF	1.07	1.63	1.98	2.15	2.48	2.75
	Proposed	1.01	1.29	1.34	1.41	1.59	1.71