Production-ready streaming anomaly detection with multiple ML algorithms and ensemble voting.
-
Multiple Algorithms
- Isolation Forest
- Statistical (Z-score)
- LSTM Autoencoder
- Ensemble Voting
-
Real-Time Processing
- Streaming data pipeline
- Low latency detection
- Configurable thresholds
-
Comprehensive Analysis
- Performance metrics
- Visualization dashboard
- Statistical evaluation
pip install -r requirements.txt
python examples/quick_demo.pyfrom src.detectors.pipeline import StreamingPipeline
# Create and train
pipeline = StreamingPipeline(detector_type='ensemble')
pipeline.train(training_data)
# Detect anomalies
results = pipeline.process_stream(test_data)
# Get anomalies
for anomaly in pipeline.anomalies:
print(f"Anomaly at {anomaly['index']}: {anomaly['value']}")src/detectors/pipeline.py(500+ lines) - Complete detection systemsrc/visualizations.py(350+ lines) - Visualization suitenotebooks/complete_demo.ipynb- Full workflowexamples/quick_demo.py- 5-minute demo
Tested on synthetic data:
- F1 Score: 0.81+
- Precision: 0.85+
- Recall: 0.78+
- Streaming data processing
- Ensemble machine learning methods
- Production system design
- Trade-offs: accuracy vs latency
Contact: Mike Ichikawa - projects.ichikawa@gmail.com