Add alert system folder with AE model, classifier and notebooks

[Samridh-Mahajan]

Description

This PR adds the Alert System module including data preparation, anomaly detection (Autoencoder), and behavioural risk classifier. It introduces notebooks, scripts, trained models, and sample output files that will be used in the Guardian project for health risk detection.

Todos

Tested and working locally in Colab + VS Code

Code follows the style guidelines of this project

Performed a self-review of all scripts and notebook flow

Documentation added for model training and usage

Requesting review from at least 2 devs (backend + AI team recommended)

How to Test

Pull this branch:

git checkout feature/alert-system

Navigate to Alert_System/.

Run the notebook Alert_System.ipynb in Google Colab or VS Code (with Python + Jupyter).

Verify that:

Autoencoder trains and saves model output (guardian_lstm_autoencoder.keras)

Classifier predictions are generated (guardian_classifier_predictions.csv)

Alerts are created for behavioural anomalies

Associated MS Planner Tasks

[AI Module: Vitals Anomaly Detection]

[Behavioural Risk Classifier Integration]

Known Issues

Current dataset still limited (approx. 1800 entries); performance may improve with larger data.

Need further integration with backend API for real-time alerting.

Model currently assumes sequence length = 12, may require parameter tuning.