Table 2 A brief introduction of SML methods

Techniques of ensemble learning

These techniques combine the predictions of two or more base models built with a given algorithm in order to improve overall accuracy and robustness of model

Random Forest (RF). It fits a number of decision trees classifiers on various sub-samples of training data and combines all prediction of classifiers to improve accuracy of model and solve the over-fitting problem (Breiman 2001)

XGBoost. It is an optimized distributed gradient boosting library designed to be highly efficient, flexible and portable, which implements machine learning methods under the framework of Gradient Boosting (Chen and Guestrin 2016)

LightGBM (LGBM). It is a gradient boosting framework that used decision tree-based learning algorithms. It adopts a leaf-wise tree growth strategy and introduce novel techniques including gradient-based one-side sampling and exclusive feature bundling (Ke et al. 2017)

Techniques of graph neural network

The aim of these techniques is to learn an embedding model that contains information of its neighborhood, which can be used to tackle a variety of issues, such as node and graph classification

Graph Convolutional Networks (GCN). the main idea of the GCN is to learn hidden layer representations that encode both local graph structure and features of nodes, and then the hidden layer representations will be passed through a neural network for node classification or graph classification (Kipf and Welling 2016)