Prediction
This project presents a novel approach to discovering and predicting causal relationships in streaming numerical data purely through mathematical pattern recognition, without the need for labeled datasets or explicit domain knowledge.
Key Innovation: Unlike traditional event detection systems, this system captures entire causal mechanisms — the complete sequence of changes from a period of stability through various transformations, leading to a "dramatic outcome." Each unique pattern discovered is assigned a Global Unique Identifier (GUID), essentially becoming a learned "causal law" for future predictive matching.
-
Unsupervised Causal Discovery: Automatically identifies and learns sequences of data changes that consistently precede significant events.
-
Pure Mathematical Pattern Recognition: Operates solely on statistical summaries of numerical tensors, making it highly adaptable across diverse domains (e.g., video streams, financial markets, system monitoring) without domain-specific training.
-
Complete Sequence Capture: Stores the full "causal chain" from stability to dramatic outcome, providing deep insights into event precursors.
-
GUID-based Causal Laws: Each learned pattern receives a unique identifier, making the discovered "laws" traceable and reusable for prediction.
-
Predictive Intelligence: Leverages learned causal laws to match against new, incoming data streams and predict future dramatic outcomes with a quantifiable similarity confidence.
-
Real-time Streaming Ready: Designed with efficient data structures (
deque) to handle continuous, frame-by-frame data processing.
These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.
-
Python 3.7+
-
numpy
You can install numpy using pip:
pip install numpy-
Clone the repository:
git clone [https://github.com/kmesiab/pure-causal-chain-detection.git](https://github.com/kmesiab/pure-causal-chain-detection.git) cd pure-causal-chain-detection -
Save the code: Copy the provided Python code into a file named
causal_detector.py(or any.pyfile you prefer) within the cloned directory.
The project includes a built-in demonstration function
(test_pure_causal_detection) that simulates a data stream, learns a causal
pattern, and then tests its predictive capabilities.
To run the demo:
python causal_detector.pyYou will see output similar to this:
🧮 Pure Mathematical Causal Detection System
==================================================
Discovering causal laws through mathematical pattern recognition...
Each sequence gets a unique GUID representing a causal mechanism.
Step 0: 3 objects, change=0.000 (initial)
Step 1: 3 objects, change=0.000 (stable)
Step 2: 3 objects, change=0.000 (stable)
Step 3: 3 objects, change=0.000 (stable)
Step 4: 3 objects, change=0.097 (normal)
Step 5: 3 objects, change=0.024 (stable)
Step 6: 3 objects, change=0.041 (stable)
🔍 CAUSAL SEQUENCE: cfcb711e
Length: 3 steps
Effect: 0.151
Pattern: transformation_diminishment_convergence
---
Step 7: 3 objects, change=0.088 (stable)
Step 8: 3 objects, change=0.185 (DRAMATIC)
Step 9: 6 objects, change=0.383 (DRAMATIC)
Step 10: 6 objects, change=0.515 (DRAMATIC)
Step 11: 4 objects, change=0.160 (DRAMATIC)
📊 DISCOVERY COMPLETE
System learned 1 complete causal sequences
Each GUID now represents a mathematical law for prediction.
🔍 Discovered Causal Laws:
• cfcb711e: 3 steps → effect 0.151
Pattern: transformation_diminishment_convergence
🔮 PREDICTIVE INTELLIGENCE TEST
----------------------------------------
Testing prediction on new, unseen sequence...
System will match against learned causal patterns.
🎯 CAUSAL PREDICTION SUCCESSFUL!
📋 Matching Law: cfcb711e
🎲 Similarity: 99.0%
⚡ Expected Effect: 0.151
⚠️ Prediction: Dramatic change likely
🔖 Pattern Type: transformation_diminishment_convergence
💡 The system recognized this sequence matches a known causal law!
Pure mathematical intelligence achieved predictive capability.
============================================================
🧠 PURE MATHEMATICAL CAUSAL INTELLIGENCE ACHIEVED
============================================================
✅ System discovers causal laws through pattern recognition
✅ Each law gets unique GUID for predictive matching
✅ High-confidence predictions from mathematical similarity
✅ No domain knowledge or training labels required
✅ Ready for real-world streaming data applications
Total discovered causal laws: 1
System ready for deployment! 🚀
causal_detector.py(or your chosen file name): Contains the main Python classes and logic for the causal detection system.CausalSequence: A dataclass for storing discovered causal patterns.PureCausalDetector: The core class for learning and predicting causal sequences.StreamingCausalProcessor: An interface for real-time data processing.test_pure_causal_detection(): A demonstration function.
This system's domain-agnostic nature and focus on causal sequence detection open up numerous possibilities:
- Video Stream Analysis: Predict critical events (e.g., object fragmentation, abnormal movement) in surveillance, autonomous driving, or robotics feeds.
- Behavioral Pattern Recognition: Learn and predict complex animal or human behaviors in scientific studies or interactive systems.
- Complex System Monitoring: Identify early warning signs and failure patterns in industrial machinery, IT infrastructure, or smart grids.
- Financial Market Analysis: Discover recurring price movement patterns that precede significant market shifts or asset value changes.
- Scientific Discovery: Automatically uncover novel causal relationships in large experimental datasets in biology, chemistry, or physics.
- Expand Feature Set: Investigate additional mathematical features for tensor summarization to capture more nuanced patterns (e.g., entropy, specific moment calculations).
- Advanced Sequence Alignment: Explore dynamic time warping (DTW) or other sequence alignment algorithms for more flexible similarity comparisons.
- Pattern Generalization: Implement methods to generalize or cluster similar discovered causal laws to reduce redundancy and improve robustness.
- Persistence Layer: Add functionality to save and load discovered causal sequences to/from a file or database for long-term learning.
- Visualization Tools: Develop interactive visualizations to better understand the learned sequences and the
feature_historyandchange_history. - Real-world Data Integration: Demonstrate the system's capabilities with real-world streaming datasets from various application domains.
Contributions are welcome! If you have suggestions for improvements, new features, or bug fixes, please feel free to open an issue or submit a pull request. Please read our Code of Conduct before contributing.
If you discover any security-related issues, please do not open a public issue. Instead, please contact kmesiab directly via email. We appreciate your responsible disclosure.
This project is licensed under the MIT License - see the LICENSE file for details.
- Inspired by the pursuit of fundamental intelligence in complex systems.
- Built with the power of
numpyand standard Python libraries.
This project operationalizes concepts discussed in the following articles:
- Emergent Concept Modeling: A Paradigm Shift in AI
- The Language Myth: Rediscovering Cognition
- Beyond Words: Toward Multidimensional Cognition
For any questions, collaboration opportunities, or further information, please reach out to:
- GitHub: @kmesiab
- LinkedIn: Kevin Mesiab
This project was developed by kmesiab, and it operationalizes the concepts discussed in the linked articles.