Skip to main content

Table 5 Random forest model performance with undersampling

From: Detecting DeFi securities violations from token smart contract code

 

Accuracy

Precision

Recall

F-1 score

RF1: Full-feature model

0.759

0.757

0.759

0.757

RF2: Top 10 features of model RF1

0.801

0.800

0.800

0.800

RF3: Top 3 features of RF1

0.780

0.780

0.780

0.780

RF4: Top feature of RF1 (CALLDATASIZE)

0.777

0.774

0.777

0.774

RF5: Non-zero coefficients of EN regression model

0.731

0.731

0.731

0.731

RF6: Top 10 features of RF5

0.743

0.742

0.743

0.742

RF7: Top 3 features of RF5

0.741

0.741

0.741

0.741

RF8: Top feature of RF5 (LT)

0.747

0.747

0.747

0.747

RF9: Top 10 features of EN regression model

0.690

0.689

0.690

0.689

  1. The final model is indicated in bold