Meta-CoSTNet

Meta-learning Multiple Transportation Pattern for Traffic Demand Prediction

Getting Started

Data

Please refer to the files under preprocess/.

Environment

conda create -n $ENV_NAME$ python=3.7
conda activate $ENV_NAME$

# Install compatible PyTorch 1.11.0, for example, for CUDA 11.3
pip install torch==1.11.0+cu113 --extra-index-url https://download.pytorch.org/whl/cu113 
# Or, CUDA 10.2 
pip install torch==1.11.0+cu102 --extra-index-url https://download.pytorch.org/whl/cu102 

pip install -r requirements.txt

Saved models

You can download our trained meta auto-encoder models and demand prediction model. Place them under the directory results/saved_models/

costnet/ 
    results/saved_models/
        MLAEN_1/
            bike_drop.pth
            bike_pickup.pth
            taxi_drop.pth
            taxi_pickup.pth
        Ours_best/
            best.pth
    preprocess/
    costnet/
        ...

Run models

Train our model

python train.py --model AutoEncoder --trainer MLAE --ddir ../datasets/ --dname multi-nyc-ae --config MLAE_config
python train.py --model Ours --trainer CoSTNet --ddir ../datasets/ --dname multi-nyc-lstm --config Ours_config

You can also train the baseline models, by just changing the arguments model. Delte the argument trainer. Similarly, the code loads $MODEL_NAME$_config.py by default. Below lists the available baseline models.

• MLP
• ConvLSTM
• STResNet

Test our model

python test.py --model Ours --trainer CoSTNet --ddir ../datasets/ --dname multi-nyc-lstm --config Ours_config --ckpt ../results/saved_models/Ours_best/best.pth

Similarly, you can also test the baseline models, by just changing the arguments.

Train and test CoSTNet

python train.py --model AutoEncoder --ddir ../datasets/ --dname multi-nyc-ae 
python train.py --model CoSTNet --ddir ../datasets/ --dname multi-nyc-lstm 
python test.py --model CoSTNet --ddir ../datasets/ --dname multi-nyc-lstm --ckpt $CKPT_PATH$