Table 13 Ablation study of proposed method.

From: An elegant intellectual engine towards automation of blockchain smart contract vulnerability detection

Methods

Accuracy (%)

Precision (%)

Recall (%)

F1-score (%)

Edge attention-residual graph convolutional networks (EA-RGCN)27

90.47

91.16

89

90.03

BERT34

92.53

94.21

95.77

89.47

Adaptive neuro-fuzzy inference system (ANFIS)26

96.4

99.4

96

99.1

CNN-LSTM43

91

92

99

95

HB3LSTM + explainable AI (proposed)

99.68

99.43

99.54

99.40