Skip to main content

Table 6 Feature importance for random forest models with undersampling

From: Detecting DeFi securities violations from token smart contract code

RF1

RF2

RF3

RF5

RF6

RF7

RF9

Feature

Imp.

Feature

Imp.

Feature

Imp.

Feature

Imp.

Feature

Imp.

Feature

Imp.

Feature

Imp.

CALLDATASIZE

0.062

CALLDATASIZE

0.203

CALLDATASIZE

0.340

LT

0.077

LT

0.165

LT

0.357

SWAP2

0.206

LT

0.034

LT

0.125

CALLVALUE

0.334

CALLVALUE

0.075

CALLVALUE

0.158

CALLVALUE

0.340

DUP2

0.192

CALLVALUE

0.032

CALLVALUE

0.123

LT

0.326

CALLER

0.046

CALLER

0.111

CALLER

0.303

MLOAD

0.180

SHR

0.029

SWAP3

0.118

  

DUP2

0.043

SWAP2

0.094

  

RETURN

0.152

EXP

0.026

EXP

0.110

  

SWAP2

0.043

DUP2

0.094

  

SHL

0.122

SWAP3

0.026

CALLER

0.089

  

MLOAD

0.039

MLOAD

0.088

  

PUSH21

0.061

NUMBER

0.023

SHR

0.072

  

DUP7

0.039

DUP7

0.084

  

SLT

0.031

PUSH5

0.021

NUMBER

0.063

  

DUP5

0.033

DUP5

0.074

  

LOG4

0.020

CALLER

0.018

PUSH5

0.055

  

GT

0.033

GT

0.071

  

PUSH30

0.019

ADDRESS

0.018

ADDRESS

0.041

  

DUP11

0.032

DUP11

0.061

  

PUSH9

0.018