Skip to main content

Architecture

Author: Cansu Moran

Pipeline Overview​

The reservoir computing system comprises several components designed to streamline data preprocessing, prediction, and storage. Below is an outline of the architecture:

  1. Data Ingestion

    • Historical FCS and RCSI data for each country are fetched daily.
    • New data is used for the predictions, ensuring the model operates on the latest information.

  2. Preprocessing

    • Data is structured for compatibility with the reservoir computing model.

  3. Reservoir Computing Model

    • Built using the scan package from DLR.
    • Generates predictions for FCS and RCSI values, along with their bounds, for the next 90 days.

  4. Prediction Update and Storage

    • Predictions are rounded to the nearest integer.
    • Data is saved to the database, with unique keys (adm0_code, prediction_date) ensuring accurate updates and insertions.

  5. API Endpoint for Predictions

    • Predictions are exposed through an API endpoint that provides structured data for frontend visualization.

Diagram​

[ Data Ingestion ] ---> [ Preprocessing ] ---> [ Reservoir Computing Model ] ---> [ Database ] ---> [ API Endpoint ]