Shared Mobility Demand Forecasting

• Shared Mobility Demand Forecasting
  • 1. Installation
    • 1.1. Clone the repository:
    • 1.2. Create and run Docker container:
    • 1.3. Specify Parameters for the Trainer:
  • 2. Example Usage:
• Acknowledgments

1. Installation

1.1. Clone the repository:

git clone https://github.com/DataStories-UniPi/Shared-Mobility.git
cd Shared-Mobility

1.2. Create and run Docker container:

To ensure compatibility, we provide a Dockerfile to run the model inside a container:

docker build -t shared_mobility .
docker run -it shared_mobility /bin/bash

NOTE: If you plan to run this in a linux environment, you need to specify the platform flag --platform linux/x86_64 during the build

1.3. Specify Parameters for the Trainer:

In order to centralize the training process we have created a so-called Orchestrator. The Area-of-Interest, model variation and device can be selected via the following arguments:

-h, --help                      show this help message and exit
--city {Amsterdam, Rotterdam}   Select city to run experiment on                     
--method {reg, classif}         Select the model variation
--device {cpu, cuda}            Select whether you want to use CPU or GPU

2. Example Usage:

In order to train the Regressor model on Rotterdam using CPU, run the following command:

 python src/orchestrator.py --city Rotterdam --method reg --device cpu 

Acknowledgments

The research work was supported by the Horizon Europe R&I programme EMERALDS under the GA No. 101093051.