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

AI Guardian/AlertSystemGMAI/README.md

@@ -0,0 +1,43 @@
+# Alert System Task – William Jackson
+
+## Overview
+
+Behavioural anomaly detection pipeline using an LSTM autoencoder and simple classifiers.  
+The autoencoder is trained on supporter behaviour sequences and reconstruction error is used to map anomalies into Low / Medium / High risk.  
+Random Forest and MLP classifiers are also trained to detect anomalies from the same features.
+
+## Features
+
+- Steps taken
+- Calorie intake
+- Sleep hours
+- Water intake (ml)
+- Bathroom visits
+- Exercise minutes (if available)
+- Meals skipped (if available)
+
+## Contents
+
+- `AlertSystemScript.ipynb` – main notebook (training, scoring, exporting results)
+- `artifacts/` – models and outputs
+  - `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` – shortlist of medium/high alerts top 30
+
+## How to run (Colab)
+
+1. Upload the dataset with original headers (`patientId, observationEnd, stepsTaken, calorieIntake, sleepHours, waterIntakeMl, bathroomVisits, exerciseMinutes, mealsSkipped`) to `/content/`.
+2. Set the dataset path at the top of the notebook:
+   ```python
+   CSV_PATH = "/content/New AI spreadsheet - Sheet1.csv"
+   ```
+3. Run the notebook end-to-end.
+4. Outputs will be saved to /content/:
+
+- guardian_ae_predictions.csv
+- guardian_classifier_predictions.csv
+- top_alerts.csv

AI Guardian/AlertSystemGMAI/RUNBOOK_WJ_Alerts.md

@@ -0,0 +1,44 @@
+# Runbook – William Jackson
+
+## Requirements
+
+- Python 3.9+
+- Recommended environment: Google Colab
+- Packages needed: numpy, pandas, scikit-learn, tensorflow, matplotlib, joblib
+
+In Colab install with:
+
+```bash
+!pip install numpy pandas scikit-learn tensorflow matplotlib joblib
+```
+
+## Dataset
+
+File: New AI spreadsheet - Sheet1.csv  
+Headers required:  
+patientId, observationEnd, stepsTaken, calorieIntake, sleepHours, waterIntakeMl, bathroomVisits, exerciseMinutes, mealsSkipped
+
+Upload the CSV to /content/ when running in Colab.
+
+## Steps
+
+1. Upload dataset to /content/.
+2. Open AlertSystemScript.ipynb.
+3. At the top, set the dataset path:
+   ```python
+   CSV_PATH = "/content/New AI spreadsheet - Sheet1.csv"
+   ```
+4. Run the notebook in its entirety
+5. Outputs will be saved to /content/:
+   - guardian_ae_predictions.csv
+   - guardian_classifier_predictions.csv
+   - top_alerts.csv (top 30 medium/high alerts)
+
+## Artifacts
+
+The ipynb file Training also creates:
+
+- guardian_lstm_autoencoder.keras
+- guardian_behavior_rf.joblib
+- guardian_behavior_mlp.joblib
+- scaler.npy