HyperD

A pytorch implementation for the paper: "HyperD: Hybrid Periodicity Decoupling Framework for Traffic Forecasting".

Getting Started

Environment Requirements

The code is built based on Python 3.12.7, PyTorch 2.5.1, and EasyTorch.

We implement our code based on BasicTS.

Dependency can be installed using the following command:

pip install -r requirements.txt

Download Data

You can download data from BasicTS and unzip it into the datasets directory.

We use four datasets: PEMS03, PEMS04, PEMS07, and PEMS08.
Each dataset directory should contain the following three files:

• data.dat: time series data
• adj_mx.pkl: adjacency matrix
• desc.json: dataset description and metadata

Statistical Prior Initialization

Run Initialization.py to generate the initialization matrices for learnable daily and weekly embeddings. The results will be automatically saved in the datasets/{dataset_name} directory.

For example:

python Initialization.py -d PEMS08

Two .npy files will be generated:

• daily_init.npy
• weekly_init.npy

These files will be automatically loaded during model training.

Run

To train the model, run:

python train.py --cfg='baselines/HyperD/PEMS03.py'
python train.py --cfg='baselines/HyperD/PEMS04.py'
python train.py --cfg='baselines/HyperD/PEMS07.py'
python train.py --cfg='baselines/HyperD/PEMS08.py'

Main Results