DM2T

This repository contains the source code and dataset for the paper: "Improving Traffic Load Prediction with Multi-modality: A Case Study of Brisbane".

Dataset

Datasets used for developing models are provided under /dataset. This dataset is the traffic load of the Coronation Drive in Brisbane, Australia. For information about the meaning of the column, please refer to this website.

The data provided is pre-processed. Please contact author if you need the raw data.

You can download raw data from this website by using API. Note the data is in real-time.

tweet_all_brisbane.csv file contains the data retrieved from twitter used in this paper.

Files

1. main-models-training.ipynb: main models development (CNN, RNN and LSTM)
2. sentiment-classification-for-tweets-all-brisbane.ipynb: Sentiment classification for Brisbane tweets
3. training-sentiment-analysis-model.ipynb: Training sentiment classifier
4. visualizing-main-models.ipynb: Visualizing the prediction from 3 main models
5. scraping-twitter.ipynb: Collecting twitter data

Pre-trained model

We also provide pre-trained model used in our paper under folder /trained models.

Environments detail

CUDA Version: 10.2

Nvidia GeForce RTX 2080 GPU

Python 3.8.10

Tensorflow 2.4.0

Citation

@inproceedings{tran2022improving,
    title={Improving Traffic Load Prediction with Multi-modality},
    author={Tran, Khai Phan and Chen, Weitong and Xu, Miao},
    booktitle={Australasian Joint Conference on Artificial Intelligence},
    pages={254--266},
    year={2022},
    organization={Springer}
  }