Dataset For Task

In this section, we introduce to you the correspondence between our provided standard datasets tasks and models. It is worth noting that some datasets can only support some models in a certain task.

If you think this table does not look convenient enough, please click here to view this table on Github

Task	Dataset	Supported Model	Remark
Traffic Flow Prediction	TAXIBJ, PORTO, NYCTAXI_GRID, NYCBIKE, AUSTINRIDE, BIKEDC, BIKECHI, NYCBike20140409, NYCBike20160708, NYCBike20160809, NYCTaxi20140112, NYCTaxi20150103, NYCTaxi20160102, T_DRIVE20150206, T_DRIVE_SMALL	ACFM, STResNet, DSAN, ACFMCommon, STResNetCommon	Grid based dataset
	PEMSD3, PeMSD4, PeMSD7, PeMSD8, BEIJING_SUBWAY, M_DENSE, SHMETRO, HZMETRO, NYCTAXI_DYNA	RNN, Seq2Seq, FNN, AutoEncoder, AGCRN, ASTGCNCommon, MSTGCNCommon, STSGCN, CONVGCNCommon, ToGCN, MultiSTGCnetCommon, STNN, ASTGCN, MSTGCN, CONVGCN, DGCN, ResLSTM, MultiSTGCnet	Point based dataset
	TAXIBJ, PORTO, NYCTAXI_GRID, NYCBIKE, AUSTINRIDE, BIKEDC, BIKECHI, NYCBike20140409, NYCBike20160708, NYCBike20160809, NYCTaxi20140112, NYCTaxi20150103, NYCTaxi20160102, T_DRIVE20150206, T_DRIVE_SMALL	RNN, Seq2Seq, FNN, AutoEncoder, AGCRN, ASTGCNCommon, MSTGCNCommon, STSGCN, CONVGCNCommon, ToGCN, MultiSTGCnetCommon, STNN, ASTGCN, MSTGCN, CONVGCN, DGCN, ResLSTM, MultiSTGCnet	Need simple modification for grid based dataset. See Note 3.
	M_DENSE	CRANN	Need .ext file
	NYCBike20160708, NYCTaxi20150103	STDN	Need .gridod file
Traffic Speed Prediction	METR_LA, LOS_LOOP, PeMSD4, PeMSD8, PEMSD7(M), PEMS_BAY, LOS_LOOP_SMALL, SZ_TAXI, LOOP_SEATTLE, Q_TRAFFIC, ROTTERDAM	RNN, Seq2Seq, FNN, AutoEncoder, DCRNN, STGCN, GWNET, MTGNN, STMGAT, TGCN, ATDM, HGCN, DKFN, STTN, GTS, GMAN, STAGGCN, TGCLSTM	Point based dataset.
	LOOP_SEATTLE	TGCLSTM	See Note 2.
On-Demand Service Prediction	TAXIBJ, PORTO, NYCTAXI_GRID, NYCBIKE, AUSTINRIDE, BIKEDC, BIKECHI, NYCBike20140409, NYCBike20160708, NYCBike20160809, NYCTaxi20140112, NYCTaxi20150103, NYCTaxi20160102, T_DRIVE20150206, T_DRIVE_SMALL	DMVSTNet	Grid based dataset
	PEMSD3, PeMSD4, PeMSD7, PeMSD8, BEIJING_SUBWAY, M_DENSE, SHMETRO, HZMETRO, NYCTAXI_DYNA	CCRNN, STG2Seq	Point based dataset
	TAXIBJ, PORTO, NYCTAXI_GRID, NYCBIKE, AUSTINRIDE, BIKEDC, BIKECHI, NYCBike20140409, NYCBike20160708, NYCBike20160809, NYCTaxi20140112, NYCTaxi20150103, NYCTaxi20160102, T_DRIVE20150206, T_DRIVE_SMALL	RNN, Seq2Seq, FNN, AutoEncoder, CCRNN, STG2Seq	Need simple modification for grid based dataset. See Note 3.
Od Matrix Prediction	NYCTAXI_OD	GEML	Point-OD based dataset
	NYC_TOD	GEML	Need simple modification for grid-od based dataset. See Note 4.
	NYC_TOD	CSTN	Grid-OD based dataset and .ext file
Traffic Accidents Prediction	NYC_RISK, CHICAGO_RISK	GSNet	Grid based traffic accidents dataset
Trajectory Next-Location Prediction	Gowalla, BrightKite	FPMC, RNN, ST-RNN, ATST-LSTM, DeepMove, HST-LSTM, LSTPM, STAN	Trajectory based dataset
	Fousquare, Instagram	FPMC, RNN, ST-RNN, ATST-LSTM, DeepMove, HST-LSTM, LSTPM, GeoSAN, STAN, SERM, CARA	Trajectory based dataset
Estimated Time of Arrival	Chengdu_Taxi_Sample1	DeepTTE	Trajectory based dataset
	Beijing_Taxi_Sample	TTPNet, DeepTTE	Trajectory based dataset
Map Matching	Seattle, global	STMatching, IVMM, HMMM	Trajectory based dataset
Road Network Representation Learning	bj_roadmap_edge	ChebConv, LINE	Road network dataset

Note 1

The bolded dataset is the one we recommend.

Note 2

For TGCLSTM, need to set dataset_class to TrafficStatePointDataset. Otherwise, the default dataset_class=TGCLSTMDataset is only suitable for dataset LOOP_SEATTLE.

Note 3

Here is how to generalize models used for point-based data for grid-based data.

(1) If the dataset class used by the model is TrafficStatePointDataset, such as AGCRN, ASTGCNCommon, CCRNN, etc., you can directly set dataset_class to TrafficStateGridDataset in task_file.json or through a custom configuration file(--config_file). Then set the parameter use_row_column of TrafficStateGridDataset to False.

(2) If the dataset class used by the model is the subclass of TrafficStatePointDataset, such as ASTGCNDataset, CONVGCNDataset, STG2SeqDataset, etc., you can modify the file of the dataset class to make it inherit TrafficStateGridDataset instead of the current TrafficStatePointDataset. Then set the parameter use_row_column in the function __init__() to False.

Example (1):

Before modification:

# task_config.json
"RNN": {
    "dataset_class": "TrafficStatePointDataset",
},
# TrafficStateGridDataset.json
{
  "use_row_column": true
}

After modification:

# task_config.json
"RNN": {
    "dataset_class": "TrafficStateGridDataset",
},
# TrafficStateGridDataset.json
{
  "use_row_column": false
}

Example (2):：

Before modification:

# task_config.json
"STG2Seq": {
    "dataset_class": "STG2SeqDataset",
},
# STG2SeqDataset.json
{
  "use_row_column": false
}
# stg2seq_dataset.py
from libcity.data.dataset import TrafficStatePointDataset
class STG2SeqDataset(TrafficStatePointDataset):
    def __init__(self, config):
        super().__init__(config)
        pass

After modification:

# task_config.json
"STG2Seq": {
    "dataset_class": "STG2SeqDataset",
},
# STG2SeqDataset.json
{
  "use_row_column": false
}
# stg2seq_dataset.py
from libcity.data.dataset import TrafficStateGridDataset
class STG2SeqDataset(TrafficStateGridDataset):
    def __init__(self, config):
        super().__init__(config)
        self.use_row_column = False
        pass

Note 4

Here is how to generalize models used for point-based data for grid-based data. (Similar to Note 3)

(1) If the dataset class used by the model is TrafficStateOdDataset, such as GEML etc., you can directly set dataset_class to TrafficStateGridOdDataset in task_file.json or through a custom configuration file(--config_file). Then set the parameter use_row_column of TrafficStateGridOdDataset to False.

(2) If the dataset class used by the model is the subclass of TrafficStateOdDataset, you can modify the file of the dataset class to make it inherit TrafficStateGridOdDataset instead of the current TrafficStateOdDataset. Then set the parameter use_row_column in the function __init__() to False.