Skip to main content

Table 2 Comparisons of seven classification methods on TCGA data

From: The application of sparse estimation of covariance matrix to quadratic discriminant analysis

Methods

LUSC

LUAD

TC

DLDA

0.013 (0.007, 50)

0.035 (0.027, 50)

0.1 (0.042, 50)

DQDA

0.008 (0.006,50)

0.018 (0.019, 50)

0.08 (0.047, 50)

NN

0.014 (0.01, 50)

0.027 (0.016, 50)

0.085 (0.053, 50)

SVM

0.01 (0.007, 50)

0.024 (0.017, 50)

0.088 (0.041, 50)

SCRDA

0.039 (0.031, 128)

0.044 (0.026, 95)

0.122 (0.068, 502)

RF

0.007 (0.002, NA)

0.018 (0.009, NA)

0.08 (0.039, NA)

SQDA

0.003 (0.003, 1900)

0.011 (0.009, 1900)

0.036 (0.021, 2900)

DLDA2

0.017 (0.005, 12100)

0.031 (0.013, 8900)

0.114 (0.038, 2300)

DQDA2

0.008 (0.004,10000)

0.023 (0.008, 8300)

0.107 (0.05, 5800)

Methods

PC

HNC

LC

DLDA

0.125 (0.024, 50)

0.034 (0.012, 50)

0.055 (0.017, 50)

DQDA

0.11 (0.022, 50)

0.03 (0.016, 50)

0.045 (0.021, 50)

NN

0.094 (0.029, 50)

0.032 (0.013, 50)

0.051(0.015, 50)

SVM

0.116 (0.031, 150)

0.037 (0.023, 50)

0.04 (0.014, 50)

SCRDA

0.094 (0.037, 1989)

0.039 (0.021, 2200)

0.069 (0.026, 56)

RF

0.11 (0.013, NA)

0.033 (0.013, NA)

0.048 (0.018, NA)

SQDA

0.206 (0.134, 1300)

0.021 (0.015, 2200)

0.04 (0.041, 500)

DLDA2

0.128 (0.026, 3400)

0.033 (0.01, 6600)

0.068 (0.02, 7800)

DQDA2

0.205 (0.066, 3100)

0.049 (0.022, 5900)

0.089 (0.027, 6100)

Methods

BC

KC

CC

DLDA

0.035 (0.017, 50)

0.028 (0.018, 50)

0.006 (0.008, 50)

DQDA

0.018 (0.009, 50)

0.037 (0.03, 50)

0.004 (0.006, 50)

NN

0.021 (0.013, 50)

0.031 (0.019, 50)

0.005 (0.009, 50)

SVM

0.018 (0.012, 50)

0.028 (0.018, 50)

0.004 (0.006, 50)

SCRDA

0.045 (0.019, 452)

0.047 (0.011, 78)

0.023 (0.014, 49)

RF

0.027 (0.013, NA)

0.025 (0.014, NA)

0.011 (0.011, NA)

SQDA

0.021 (0.008, 2800)

0.009 (0.005, 6100)

0.007 (0.008, 5900)

DLDA2

0.036 (0.015, 8000)

0.039 (0.007, 10200)

0.02 (0.014, 11700)

DQDA2

0.069 (0.033, 7400)

0.045 (0.035, 9600)

0.021 (0.014, 10200)

  1. The reported numbers in each table entry in the form of a (b,c) mean: a is the average prediction error, b is the standard deviation, and c is the median number of predictors selected.