Research Article

mRMR-ABC: A Hybrid Gene Selection Algorithm for Cancer Classification Using Microarray Gene Expression Profiling

Table 1

Comprehensive comparison between the state-of-the-art machine learning cancer classification methods in terms of classification accuracy and number of selected genes for the four benchmark microarray datasets (colon, leukemia, lung, and prostate). The number between parenthesis denotes the number of selected genes. The best classification performance in each gene selection approach for each microarray dataset is indicated in bold.

Cancer classification methods Colon LeukemiaLung Prostate

ANN [42] 93.43 (40)97.33 (10)
NB [20] 88.79 (8)100 (8)98.04 (8)
KNN [20]77.42 (12) 100 (12)97.06 (12)
KNN [43] 100 (9)
KNN [44] 100 (9)
RF [45]84.4 (14)91.3 (2)93.9 (18)
SVM [46]94 (20)
SVM [26]88.41 (25) 99.63 (5)90.26 (4)
SVM [47] 88.18 (95) 96.88 (88)99.90 (29) 93.41 (85)
SVM [20]91.68 (78)98.35 (37)98.29  ()
SVM [27] 100  () 100  ()
SVM [48] 91.67 (4)
SVM [49]100 (6) 98.66 (2)
SVM [50] 98.57 (7)
SVM [21] 95 (5)100  ()100  ()
SVM [28] 99.41 (10)100 (25)