Table 3.
Summary of deep learning models used in the studies reviewed with the graph-format input
| Study | Proposed/adopted model(s) | Baselines | Evaluation metrics |
|---|---|---|---|
| Year 2018 | |||
| [7] | GraphCNN-Bike (GCN+LSTM) | HA, ARIMA, SARIMA, GBRT, LSTM | RMSE |
| [41] | GCNN-DDGF (GCN) | XGBoost, LSTM, MLP, SVR, LASSO, HA | RMSE, MAE, |
| Year 2019 | |||
| [24] | BikeNet (GCN+GRU) | ARIMA, SVR, FFNN, LSTM | RMSE, MAE, MAPE |
| [35] | TGNet (GN+Temporal Guided Embedding) | ARIMA, XGBoost, ST-ResNet, DMVST-Net, STDN | RMSE, MAPE, Parameter Number |
| [2] | STG2Seq (GCN+Attention) | HA, OLR, XGBoost, DeepST, ResST-Net, DMVST-Net, ConvLSTM, FCL-Net, FlowFLexDP, DCRNN, STGCN | RMSE, MAE, MAPE |
| [40] | STG2Vec+LSTM | HA, LASSO, kNN, RF, GBRT, RNN, GRU | RMSE, MAE |
| Year 2020 | |||
| [88] | ST-CGA (GAT+CNN) | ARIMA, SVR, Fuzzy+NN, RNN, LSTM, DeepST, ST-ResNet, DMVST-Net, STDN, UrbanFM, ST-MetaNet, ST-GCN, ST-MGCN | RMSE, MAPE |
| [69] | SCEG (GCN) | GRU, T-GCN, E-GCN, Multi-graph, CG-GCN | MAPE, RMSPE |
| [18] | DTCNN (GCN+GRU) | HA, VAR, XGBoost, RNN, LSTM, GRU, DCRNN | RMSE, PCC, MAE |
| [58] | MVGCN (GCN) | HA, VAR, GBRT, FC-LSTM, GCN, DCRNN, FCCF, ST-MGCN | RMSE, MAE |
| [25] | GBikes (GAT+GCN+Attention) | HA, SHA, ARIMA, ANN, LSTM, RNN, STCNN, GC, MGN | RMSE |
| [46] | AGSTN (GCN+Attention+LSTM) | ARIMA, SVR, FC-LSTM, DCRNN, AST-GCN, ST-MGCN | MAE, RMSE, P@5, NDCG |
| [77] | BikeGAAN (GCN+Attention+LSTM) | SES, MLP, ARIMA, HA, RNN, GRU, LSTM, CNN, CNN-RNN, CNN-LSTM, CNN-GRU, GCN | MSE |
| [53] | GCN | HA, ARIMA, LSTM, DCRNN, STGCN | RMSE |
| Year 2021 | |||
| [80] | GCN+GRU+Attention | XGBoost, FC-LSTM, DCRNN, STGCN, STG2Seq, Graph WaveNet | RMSE, MAE, PCC |
| [89] | ST-GDN (Attention+GAT+GCN) | ARIMA, SVR, Fuzzy NN, ST-RNN, D-LSTM, DeepST, ST-ResNet, DMVST-Net, STDN, UrbanFM, ST-MetaNet, DCRNN, ST-GCN, ST-MGCN, GMAN | RMSE, MAPE |
| [52] | GCN+LSTM | ARIMA, SVR, LSTM, DCRNN, STGCN, T-GCN | RMSE, MAE |
| [82] | FGST (GCN+LSTM) | FNN, LSTM, GRU, GCN | RMSE, MAE |
| [12] | GCN+TCN | HA, ARIMA, ETS, RF | RMSE, MAE |
| [47] | LSGC-LSTM (GCN+LSTM) | RNN, LSTM, GRU, GAT-LSTM, AGCRN, DGCNN | SMAPE, RMSE, MAE |
| [74] | STGCN (GCN+TCN) | RNN, LSTM, GRU | SMAPE, RMSE, MAE |