https://github.com/heraclitus0/qsi
A real-time rupture detection tool for identifying forecast vs. actual drift and preventable losses.
https://github.com/heraclitus0/qsi
adaptive-thresholding control-systems drift-detection real-time-analysis rupture-detection streamlit
Last synced: about 2 months ago
JSON representation
A real-time rupture detection tool for identifying forecast vs. actual drift and preventable losses.
- Host: GitHub
- URL: https://github.com/heraclitus0/qsi
- Owner: heraclitus0
- License: mit
- Created: 2025-07-09T05:10:32.000Z (12 months ago)
- Default Branch: main
- Last Pushed: 2025-09-08T09:30:32.000Z (10 months ago)
- Last Synced: 2025-09-08T09:31:48.688Z (10 months ago)
- Topics: adaptive-thresholding, control-systems, drift-detection, real-time-analysis, rupture-detection, streamlit
- Language: Python
- Homepage: https://rupture-detector-vxcv8twev4y3vcuqzjprnw.streamlit.app/
- Size: 798 KB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE.txt
- Notice: NOTICE.txt
Awesome Lists containing this project
README
Quantitative Stochastic Intelligence
Adaptive rupture detection and epistemic diagnostics for dynamic systems.
Policy-calibrated intelligence that learns from volatility.
---
## Overview
QSI is a **decision-intelligence engine** that detects ruptures in forecast vs. actual performance, quantifies preventable losses, and provides **epistemic diagnostics** such as drift, threshold breaches, stability scope, and policy sensitivity.
It is designed for **board-level clarity** and **field-level adaptability**, aligning with volatile domains ranging from **supply chains** to **finance, cyber, and pharma**.
Minimal, calibrated, and transparent — QSI surfaces actionable intelligence without black-box opacity.
---
## Features
- **Rupture Detection** — Tracks forecast vs. actual drift, thresholds, and breach events.
- **Loss Quantification** — Converts drifts into monetary loss using unit cost.
- **Epistemic Diagnostics** — Scope score, PSI, and breach ETA forecasting.
- **Cognize Meta-Policy** — Optional adaptive mode with exploration and policy promotion.
- **Segment Graphs** — Coupled dynamics across multiple SKUs or regions.
- **Dynamic Configurability** — Every knob is exposed for user calibration, no statics hard-coded.
- **Custom Models** — Plug in enterprise-specific threshold policies via registry.
---
## Interface
---
## Use Cases
QSI is **domain-agnostic**. Example applications include:
- **Supply Chains** — Prevent procurement losses by catching over/under-forecast drifts early.
- **Finance** — Stress-test trading strategies against volatility thresholds.
- **Healthcare & Pharma** — Detect demand misalignments in critical drug or equipment supply.
- **Cybersecurity** — Monitor deviations in expected traffic or anomaly baselines.
- **Operations & Strategy** — Track policy adherence, systemic drift, and rupture clusters.
---
## Quick Start
1. **Try it live**: [Launch QSI on Streamlit](https://zkvyksd6zuzfyaqshzphzm.streamlit.app/)
2. **Upload your data**: CSV with columns → `Date, Forecast, Actual, Unit_Cost`.
3. **Explore outputs**:
- Ruptures flagged with drift × cost loss quantification
- Policy-calibrated thresholds & diagnostics (Scope, PSI, ETA)
- Interactive drift vs threshold plots & volatility bands
4. **Dive deeper**:
- [Case Study — Hyderabad Pilot](QSI_case_study.md)
- [Full Analytical Report](QSI_project_report.md)
- [User Guide](USER_GUIDE.md)
---
## Disclaimer
QSI outputs are **calibration-dependent**.
The toggles and parameters exist for a reason: to adapt the system to the volatility profile of your domain.
Misuse without domain calibration may lead to misleading results.
---