Table 1 Blockchain phishing research comparison table

Manual

After extracting features manually, classification is performed using a machine learning classification model

Advantages:

1. Can result in better categorization

Disadvantages:

1. Can lead to loss of important information

2. Requires some experience

Automatic

After learning the features using the deep learning model, the features are fed into the classification model for classification

Advantages:

1. No need to extract features manually, reducing the interference of human factors

2. Doesn't require much experience

Disadvantages:

1. Classification effect is lower than manual extraction of features

Manual + automatic

(CT-GCN+)

After extracting the features manually, they are fed into the GCN to learn the features automatically, and the graph balancing process is considered in the GCN

Combining manual and automatic feature extraction methods, while considering graph balance processing, greatly improving classification efficiency