JOR-Bayesian Fusion Framework (v3.1)

Probabilistic UAP Analysis Engine & Aerospace Safety Command Center

Version: 3.1 Final Release

📌 DOI: https://doi.org/10.5281/zenodo.18088931 📦 Zenodo Record: https://zenodo.org/records/18088931 🧩 Domain Demonstrations: Predictive Maintenance, Collision Risk Analysis 📟 Operational App Interface: Collision_Risk/Safety_Command_Center.py (Streamlit Safety Command Center UI)

The James Orion Report (JOR) Bayesian Fusion framework is a probabilistic analysis system designed to evaluate Unidentified Anomalous Phenomena (UAP) using structured evidentiary scoring combined with Markov Chain Monte Carlo (MCMC) sampling via PyMC.

The system integrates qualitative observation rubrics with quantitative Bayesian-style inference to produce conservative estimates of whether a case is consistent with a non-human origin hypothesis.

---

🚀 Key Improvements in v3.1

• Unified Parameter Sync: Global constants ($K$-calibration, priors, and weights) are centrally defined in jor_fusion.py and inherited across all execution layers, ensuring consistency between CLI scoring, the Streamlit GUI, and Bayesian inference.
• Evidence-Layer Flight Modulation: Flight-related anomalies are now treated as a probabilistic modifier on physical evidence quality rather than a post-hoc score adjustment. This ensures anomaly information is incorporated at the evidentiary interpretation stage.
• Numerical Stability Improvements: Variable clipping (0.001–0.999) prevents invalid probability states, and fixed random seeds ensure reproducible MCMC sampling.
• Dynamic Scorer Logic: The interactive CLI includes conditional scoring adjustments for high-confidence observational conditions (e.g., clear daytime visibility), reducing under-weighting of high-quality data.
• Deterministic Data Exports: Implemented a targeted .round(3) truncation limit on the Uncertainty column to permanently strip binary floating-point tail artifacts (e.g., converting 0.11900000000000005 to a clean 0.119), ensuring flawless downstream spreadsheet integration.
• Syntax Bug Resolution: Optimized the dashboard's ax.errorbar data arrays by splitting nested np.vstack calculations into discrete, standalone variables, eliminating trailing parenthesis compilation panics.
• Environmental Clarification: Clarified Environmental Strong-category wording to explicitly treat clear daytime and clear nighttime atmospheric visibility equivalently. This change improves scoring consistency and reduces interpretive variance; no mathematical weighting or posterior logic changed.

---

📟 Streamlit GUI Command Center Features (Safety_Command_Center.py)

• 🌌 Kinetic Hazard Frontier: A dynamic bubble chart using a high-visibility plasma color map to plot Solid Object Probability (SOP_Mean) against Flight Modifiers (Flight_Mod). Bubble sizing automatically scales according to MCMC data uncertainty span.
• 📊 Threat Intelligence Quadrant Map: A crosshair-synced matrix mapping physical target data against strict risk metrics, highlighting critical action limits ($>0.75$ CRI).
• 🌲 Evidentiary Uncertainty Forest: An MCMC posterior uncertainty range check that isolates active target profiles with a prominent target lock overlay against background context fleets (95% Confidence Intervals).
• 🔍 Active Track Auditor & Sidebar Controls: Real-time dropdown isolation engines that immediately focus all three graphical tracking maps onto a single selected asset.

⚠️ IMPORTANT INTERACTIVE UI NOTE: To launch and run the Streamlit Command Center dashboard (Safety_Command_Center.py), you must navigate into the Collision_Risk directory and use the specialized versions of jor_fusion.py, jor_pymc_runner.py, and jor_pymc.py stored inside that specific folder. Running the dashboard utilizing the main root folder's python files will cause dependency and path alignment errors.

---

📂 Repository Structure

File / Folder	Description
Full Methodology Report	Original theoretical framework and evidentiary rubric definitions
Collision_Risk/	Dedicated UI Folder: Contains Safety_Command_Center.py and the localized backend scripts explicitly modified to drive the Command Center interface.
jor_fusion.py	Root Scorer: Interactive CLI tool for standard, standalone terminal-based scoring.
jor_pymc.py	Root Bayesian Engine: Core PyMC-based probabilistic model for standalone MCMC sampling.
jor_pymc_runner.py	Root Batch Orchestrator: Standalone terminal tool for executing multiple case evaluations.
jor_scores.csv	Outputs: Model-generated posterior estimates, means, and credible intervals for evaluated cases.

---

🔬 Domain Demonstration Projects

The JOR Bayesian Evidence Fusion architecture is being evaluated through domain adaptation demonstrations that apply evidence-based probabilistic reasoning beyond the original UAP analysis domain.

Predictive Maintenance Demonstration

Location:

Projects/Predictive_Maintenance/

This demonstration applies the JOR evidence-fusion architecture to industrial condition monitoring using simulated vibration sensor data.

Capabilities include:

• Sensor feature extraction
• Evidence representation mapping
• Recursive health-state tracking
• Adaptive baseline calibration
• Hysteresis-based anomaly alerting
• Persistent fusion-state logging

The demonstration shows how domain-specific measurements can be translated into structured evidence inputs before probabilistic state evaluation.

Documentation:

Projects/Predictive_Maintenance/README.md

---

Cloud VM Service Health Demonstration

Location:

Projects/Cloud_VM_Service_Health/

This demonstration applies the JOR evidence-fusion architecture to cloud infrastructure and service reliability monitoring using simulated telemetry inputs.

Capabilities include:

• Cloud telemetry feature extraction
• Service health evidence mapping
• Recursive health-state tracking
• Self-calibrating baseline estimation
• Per-instance gray-failure detection
• IQR-based outlier analysis
• EMA persistence tracking
• Hysteresis-based operational alerting
• Persistent fusion-state logging

The demonstration shows how operational telemetry from distributed computing environments can be converted into structured evidence inputs for recursive probabilistic health-state evaluation.

Documentation:

Projects/Cloud_VM_Service_Health/README.md

---

📊 Example Dataset

A curated dataset of 50 evaluated UAP cases using the v3.1 framework is included:

• data/v3_1/jor_uap_cases_50_v3_1.csv

This dataset contains structured inputs and model outputs, including:

• Evidence scores ($C$, $E$, $P$)
• Flight anomaly modifiers
• SOP / NHP intermediate values
• Posterior probabilities and credible intervals

It can be used to:

• Reproduce published results
• Validate model behavior
• Test parameter sensitivity and calibration

---

🧭 Version Philosophy (v3 vs v3.1)

• v3: Additive feature-weight scoring model where anomaly indicators influence posterior estimates through weighted contributions.
• v3.1: Evidence-conditioned inference model where observational reliability (e.g., flight anomalies) modifies how physical evidence is interpreted within the likelihood structure.

This distinction is critical: v3.1 does not simply re-weight outputs—it modifies how evidence is conditioned before probabilistic inference occurs.

---

🧪 The Bayesian Logic

The framework evaluates competing hypotheses:

• $H$ (Human / Prosaic Origin)
• $NH$ (Non-Human / Anomalous Origin)

Priors are initialized as: $$P(H) = 0.80$$ $$P(NH) = 0.20$$

These priors represent conservative baseline assumptions under conditions of uncertainty.

Stochastic Flight Modeling

A key component of the model is the Flight Effect, which introduces a probabilistic modulation of physical evidence strength based on observed kinematic anomalies and assumed measurement uncertainty.

This adjustment is applied at the evidence interpretation stage: $$P_{\text{Anomalous}} = \text{clip}(P \times (1 + \text{Flight Effect}), 0.0, 0.95)$$

Where the Flight Effect is modeled as a truncated distribution: $$\text{TruncatedNormal}(\mu = \text{mod}, \sigma = 0.03, \text{lower} = 0.0, \text{upper} = 0.10)$$

This formulation ensures that anomalous flight characteristics contribute proportionally to evidence strength while remaining bounded by observational uncertainty.

K-Calibration Constant (Heuristic Likelihood Proxy)

To maintain conservative inference behavior aligned with uncertainty constraints, the framework includes a calibration constant: $$P(E|H) = \min(1, 1 - \text{NHP} + K \cdot \text{SOP})$$

Where:

• $K = 0.20$ (calibration constant aligned with AARO Standards)
• $\text{SOP}$ = Solid Object Probability
• $\text{NHP}$ = Non-Human Probability (intermediate estimate)

This expression is a heuristic likelihood proxy, not a strict probabilistic identity. It is used to stabilize inference under incomplete observational data.

---

🛠️ Usage & Execution

First, ensure you have all core framework dependencies installed:

pip install pymc pytensor pandas numpy colorama streamlit matplotlib seaborn

---

⚠️ CRITICAL ARCHITECTURE NOTE: This repository contains two entirely separate, self-contained environments that use matching file names. Running a file in the root directory has absolutely nothing to do with the files inside the Collision_Risk folder. They do not share data, inputs, or outputs.

Choose one distinct path below depending on which application you want to run:

---

🔹 Option 1: Main Root Framework

Runs the primary standalone CLI scoring and Bayesian processing engine. Uses the files located strictly in the root directory.

Step 1: Ensure your terminal is in the root directory.

Step 2: Execute the root pipeline: python jor_fusion.py
python jor_pymc_runner.py

---

🔹 Option 2: Collision Risk Index Module (Collision_Risk/)

Runs the independent interactive safety dashboard and its localized backend scripts. This module operates entirely within its own directory.

Step 1: Change your terminal directory into the subfolder: cd Collision_Risk

Step 2: Launch the interactive command center dashboard: streamlit run Safety_Command_Center.py

(Note: If you need to run terminal pipeline variations for this specific safety module, use the localized jor_fusion.py or read the separate readme.md located inside this subfolder).