Skip to content

EYLeong/traffic_prediction

Repository files navigation

SUTD 50.039 Theory and Practice of Deep Learning Big Project

Traffic Congestion Prediction using Spatio-Temporal Graph Convolutional Networks

Project Summary

Our project aims to conduct traffic congestion prediction using a Spatial-Temporal Graph Convolutional Network (STGCN) trained on Singapore traffic speed dataset by Land Transport Authority (LTA). Using the STGCN, we further conducted various analyses on its effectiveness across time and across geographical regions (different roads).

Github Directory Guide

data/processed: Contains the traffic dataset used in this project

interactive-app: Code for deploying the model on the web-app

saved-models: Contains the weights for the models

Analysis.ipynb: Notebook to generate the analysis used in the Results and Analysis section of the report

Arima.ipynb: Notebook to train the alternative model used in the Comparison with Other Models section of the report

Data Visualisation.ipynb: Notebook to generate the visualisation images used in the Data Visualisations section of the report

Hyperparameter Tuning.ipynb: Notebook used to test the different hyperparameter settings for the model. The results were used and elaborated further in the Hyperparameter Tuning section of the report

STGCN Timestep Comparison.ipynb: Notebook used to test the effects of using different input and output timesteps for the model, as described in the Impact of Input and Output Timesteps section of the report

STGCN Traffic.ipynb: Main Notebook used to train the STGCN model

analysis.py: Supporting code that was used in the Analysis.ipynb notebook

model.py: Code for our STGCN model

model_utils.py: Supporting code that is used to train the model

preprocessing_utils.py: Code for preprocessing the traffic data

visualisation.py: Supporting code that was used in the Data Visualisation.ipynb notebook

About

Deep Learning approach to traffic speed prediction

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages