or4k2l/Crisis-Detector
Automated time series-based anomaly detection framework for crisis monitoring. Detects unusual behaviors, spikes, and trends for early warning systems in finance, industry, and cyber-security.
What's novel
A unified framework for detecting anomalies and crisis events across multiple domains including finance, seismology, gravitational waves, and neurophysiology. In Cooperation whith AI.
Code Analysis
10 files read · 3 roundsA domain-agnostic time-series anomaly detector that combines rolling z-scores, volatility thresholds, and Isolation Forest to flag crisis regions, with thin data loaders for finance, seismology, gravitational waves, and EEG.
Strengths
Clean, well-documented code with good separation of concerns, sensible edge-case handling (empty signals, NaN, constant signals), and a meaningful test suite that validates detection behavior with synthetic spikes and parameter sensitivity.
Weaknesses
The core algorithm is a straightforward combination of standard statistical tests and an off-the-shelf Isolation Forest with minimal feature engineering; the 'multi-domain' support is just thin wrappers around external libraries, and the crisis score formula is arbitrary with no domain-specific tuning or validation against real crisis events.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
19
Contributors
2
Files
21
Active weeks
2
Repository
Language
Python
Stars
0
Forks
0
License
MIT