Guardian Alerts: William Jackson | LSTM anomaly + RF/MLP classifier, alert logic & runbook

[jacksonwilliamyu]

Description

Implements the Alert System task – adds LSTM anomaly detection and behavioural classifiers with alert logic.

Changes in this PR

• Added AI Guardian/AlertSystemGMAI/AlertSystemScript.ipynb (main notebook for training and scoring)
• Added README.md and RUNBOOK_WJ_Alerts.md (overview + run instructions)
• Added artifacts:
  • guardian_lstm_autoencoder.keras
  • guardian_behavior_rf.joblib
  • guardian_behavior_mlp.joblib
  • scaler.npy
  • guardian_ae_predictions.csv
  • guardian_classifier_predictions.csv
  • top_alerts.csv (top 30 medium/high alerts, demo only)

How to test

1. Open AlertSystemScript.ipynb in Colab.
2. Upload New AI spreadsheet - Sheet1.csv into /content/.
3. Set the dataset path at the top of the notebook:

   CSV_PATH = "/content/New AI spreadsheet - Sheet1.csv"

4. Run all the cells in its entirety
5. Check output in /content/:
   • guardian_ae_predictions.csv
   • guardian_classifier_predictions.csv
   • top_alerts.csv

Known Issues

• n/a