Happy Valley

https://github.com/sjruan/malmcs

Paper

If you find our code or dataset useful for your research, please cite our paper:

Sijie Ruan, Jie Bao, Yuxuan Liang, Ruiyuan Li, Tianfu He, Chuishi Meng, Yanhua Li, Yingcai Wu and Yu Zheng. “Dynamic Public Resource Allocation based on Human Mobility Prediction.”, ACM IMWUT/UbiComp 2020.

Requirements

Python 3.6

  • numpy==1.14.5
  • networkx==2.2
  • shapely==1.6.4
  • pickle

Dataset

We organize our dataset into two archives, i.e., MALMCS_data.zip and PREDICTION_data.zip

  1. MALMCS_data.zip
  • frames_20180101_20181101_24.npy: this is the hourly crowd flows data in Beijing Happy Valley from 01/01/2018 to 01/11/2018 scraped from the Tencent Heat Map. The last month is used for evaluation, and previous months are used for training & validation.
  • pred_all_stresnet_mf4_masked.pkl: this is the predicted results from the prediction model for evaluation acceleration purpose. In the paper, those results are obtained by training MF-STN.
  1. PREDICTION_data.zip

This archive provides some external factors for crowd flow prediction, which can be used to train the crowd flow prediction model together with frames_20180101_20181101_24.npy. This dataset is also a data source for UrbanFM.

  • holiday features: external/holiday_20180101_20181101_24.npy
  • meteorology features: external/mete_cy_20180101_20181101_24.npy
  • ticket price features: external/price_20180101_20181101_24.npy
  • time of day features: external/tod_20180101_20181101_24.npy

Usage

Tunable Parameters

  • Service radius radius
  • Energy limitation cost_limit
  • Number of agents k

 python evaluate.py

License

The code and data are released under the MIT License.

Leave a Reply

Your email address will not be published. Required fields are marked *

Search

Popular Posts

  • LargeST
    LargeST

    A Benchmark Dataset for Large-Scale Traffic Forecasting.

  • Happy Valley
    Happy Valley

    Dynamic Public Resource Allocation based on Human Mobility Prediction.

  • LaDe
    LaDe

    The First Comprehensive Last-mile Delivery Dataset from Industry.

Categories

Archives

Tags

There’s no content to show here yet.