Overview
SARAHAI-DEMOv7.7 is a Streamlit-based demonstration showcasing:
Rolling IsolationForest for anomaly detection (CPU or GPU via cuML).
Peer-aware anomalies per U.S. Patent No. 11,308,384.
Day-ahead price simulations plus accuracy metrics.
Up to 8 stock tickers: 4 default (MSFT, AMZN, GOOGL, TSLA) + up to 4 user-input.
Running the EXE
Locate SARAHAI-DEMOv7.7.exe on your Windows 11 system.
Double-click the .exe to launch.
A console window may open to initialize the embedded Streamlit server.
The app will attempt to open your web browser at http://localhost:8501 (or another free port).
Use the browser-based interface to interact with the app:
The sidebar (left) sets parameters and toggles live updates.
The main panel (center) displays the chart, anomalies, and daily accuracy.
If EXE Fails
Some GPU drivers or dependencies might be missing or improperly bundled.
Alternate: run the original Python script if you have Python + Streamlit installed:
bash
Copy
streamlit run SARAHAI-DEMOv7.7.py
This ensures all Python-based dependencies can be controlled from your environment.
Interface Walkthrough
Sidebar:
Add Up to 4 Extra Tickers: Type any valid or test ticker (blank if not needed).
IsolationForest Config:
Contamination: portion of outliers in data.
Rolling Window: days for rolling mean/std.
Peer Window: ±days to check peer anomalies.
Pause/Resume: Toggle the day-by-day loop.
Main Panel:
Stock Logos (4 default + your extras).
Live Chart: ~180 days of “Actual” vs. “Day-Ahead Predicted” lines, plus anomaly markers.
Metrics: Each ticker’s current day’s actual vs. predicted, overall day-ahead accuracy.
Footer: Day index (0–179) & “Live Updates: Running/Paused.”
Day-by-Day Loop
Each ~1 second, the app:
Moves to next day’s data (resets at day 180),
Re-fits IsolationForest if enough data,
Marks anomalies, checks peer-based anomaly events,
Updates chart & metrics.
GPU vs. CPU
If the embedded environment includes cuML + NVIDIA drivers, a sidebar message says “Using cuML (NVIDIA).” Otherwise, “CPU-based sklearn.”
This does not guarantee your system has correct CUDA drivers for GPU usage, so proceed carefully if the .exe is missing them.
Troubleshooting
No Data: The chart might initially show “No data loaded.” Wait ~1 second.
Large Rolling Windows or many tickers can slow performance.
EXE Launch Fails: Possibly missing .dlls or GPU libs. Try the Python script route.
Summary
Double-click SARAHAI-DEMOv7.7.exe, open the link in your browser, and explore day-by-day anomalies plus day-ahead accuracy.