Mapping server and the machine learning module

Machine learning module

This module is responsible for estimating the location of the Beacon devices. It uses data from the database gathered through room calibration.
Model trained after every new room calibartion and every room deletion to ensure data consitency.

Used model

We used KNN (k-nearest neighbours) algorithm to estimate the location of the Beacon devices. The KNN algorithm is a simple and effective machine learning algorithm used for classification and regression tasks. In our implementation number of neighbours is equal 2.

Tech stack

• Python (version: 3.12.4)
• SciKit-Learn

Mapping server

This server provides communication between the main server and the machine learning module.

Tech stack

• Python (version: 3.12.4)
• Uvicorn
• FastAPI

Starting the server:

It is not recommended to run the server alone. Be sure to run the whole Deployment.

Installing the virtual environment

  pip install virtualenv

Create virtual environment (first usage)

    virtualenv .venv

Activate virtual environemnt

  source ./.venv/Scripts/activate

Get all the requriements

  pip install -r requirements.txt

Development start Python server

  uvicorn Server:app --host 0.0.0.0 --port 8083 --reload

Available Endpoints

POST "/location"

Returns the identifier of the predicted location based on the received data. If the model has not been previously trained, an error with code 503 will be returned as a response.

Arguments:

• scan_results: NetworkInfo[] - Information about networks scanned by the Beacon.

Response:

• Room identifier string.

POST "/retrain"

Allows clearing the currently saved model, and then initiates retraining using the latest information stored in the database.