MSS data forecast system

Introduction

This is a system for forecasting the Mobile Spatial Statistic (MSS) data. The system is based on the Mobaku data.

Requirements

Python 3.11.6

Usage

1. Download the MSS data and store unipped in HOME_DIR directory.

2. Create an .env file with the following contents:

   HOME_DIR = <path to the directory where the data is stored> / Folder path
   JAPAN_MESH4 = <path to the Japan Mesh data> / GeoJSON file address
   

3. In main.py modify the parameters:

   MESH_ID = <`int` with MESH4 code> (e.g. 503324732) (not in used) 
   AOI_NAME = <`str` name of project> (folder name of project)
   AOI_POLYGON = <`str` path of a `shp` polygon file to extract MSS data in mesh4.>
   AOI_MESH = <`str` path to a `geojson` file of mesh4 data.>
   EVENT_NAME = <`str` name of event>
   EVENT_DATE_START = <`date string`> (e.g. "2023-11-01")
   EVENT_DATE_MAIN = <`date string`> (e.g. "2023-11-10")
   EVENT_DATE_END = <`date string`> (e.g. "2023-11-20")
   FILE_PREFIX = <`str` prefix for files. No spaces.>
   SEASONALITY = <`int` parameter for SARIMA>
   HOURS_TO_FORECAST = <`int` hours to forecast.> (currently 3 is only possible)
   PERCENTAGE_OF_DATA_FOR_TRAINING = <`float` number to split training and test data> (e.g. 0.7)
   

4. Run python main.py to start the system.

Output

The system will output the following files: AOI_NAME is used to create the folder of the project. Inside this folder each simulation will be stored in a subfolder with the current date and time as folder name (i.e. 20231201105613 is 2023-12-01 10:56:13). Within the datetime folder, the following folders will be created:

• data : contains the data used for the simulation.
  • GeoJSON of the area of interest.
  • Pickle file with the MSS data and the Class Object.
• figures : contains the figures and plot animation of the simulation.
• output : contains a csv file with the forecasted data. (meshid,1h,2h,3h)
• plots : contains the plots of the simulation (i.e. data, forecast errors, etc.).