NMFBFS: A NMF-Based Feature Selection Method in Identifying Pivotal Clinical Symptoms of Hepatocellular Carcinoma

Table 6

Classification accuracy among three feature subsets on the training set (120 representative samples). FS₁ was obtained by the proposed approach with a given threshold (), in which 33 original symptom features and 6 new mixed features were included. FS₂ denotes the above-mentioned 33 original symptom features (). OFS indicates all the 49 symptoms before NMF calculation.

Feature subsets	Dimension	Classification accuracy in LSSVM (%)
FS1	39	80.002 ± 9.95
FS2	33	77.50 ± 12.36
OFS	49	72.50 ± 11.64