LfN_project

Repository of LFN course's project a.y. 2022/23

Paper

https://github.com/CristianBold4/LfN_project/blob/main/LFN_Final_Report.pdf

Collaborators:

• Boldrin Cristian
• Makosa Alberto
• Mosco Simone

Traffic forecasting using GCNNs

Goal: use historical speed data to predict the speed at a given future time step.

Datasets:

1. METR-LA: DCRNN author's Google Drive
2. PEMSD4-BAY: DCRNN author's Google Drive
3. PeMSD7-LA: STGCN author's GitHub repository

Data model:

Each node represents a sensor station recording the traffic speed. An edge connecting two nodes means these two sensor stations are connected on the road. The geographic diagram representing traffic speed of a region changes over time.

STGCN model

PyTorch implementation of slightly modified version of the paper Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting (https://arxiv.org/abs/1709.04875)

Requirements

To install requirements:

pip3 install -r requirements.txt

Run Program

cd STGCN-model
python main.py [--args]