How angelfdezgetino/adaptive-massive-event-detection works
This repository detects unusual public-transport demand peaks in station time series. A user selects a station and year; the pipeline preprocesses daily demand, estimates seasonal baselines and CEI/PEI behaviour, then applies an adaptive IQR or MAD detector before filtering and exporting candidate events. It can evaluate results against optional ground-truth calendars and plot single-station analyses. A separate heatmap script reads exported detections for calendar-style visualization. The workflow is documented in the README; source excerpts show pipeline internals and heatmap input handling, but not every stage’s implementation details.
Main entry points: event_detection_pipeline.py Station-Year Pipeline · heatmap.py Event Heatmap
Components 15
Pipeline
- Interactive Runner
Python/event_detection_pipeline.py - Station-Year Pipeline
Python/event_detection_pipeline.py
Detection Analysis
- Daily Preprocessing
Python/event_detection_pipeline.py - Seasonal Baseline
Python/event_detection_pipeline.py - CEI/PEI Indicators
Python/event_detection_pipeline.py - Hampel + STL
Python/event_detection_pipeline.py - Adaptive Detection
Python/event_detection_pipeline.py - Detection Filtering
Python/event_detection_pipeline.py
Evaluation & Results
- Metrics Evaluation
Python/event_detection_pipeline.py - Detected Events CSV
Visualization
- Station Plots
Python/event_detection_pipeline.py - Event Heatmap
Python/heatmap.py
Other components
- User
- Station Data
- Event Calendars
Connections 14
- User → Interactive Runner: selects station/year
README.md - Station Data → Interactive Runner: loads CSV
README.md - Interactive Runner → Station-Year Pipeline: runs station-year
README.md - Station-Year Pipeline → Daily Preprocessing: preprocesses demand
README.md - Daily Preprocessing → Seasonal Baseline: decomposes series
README.md - Seasonal Baseline → CEI/PEI Indicators: estimates behaviour
README.md - CEI/PEI Indicators → Hampel + STL: assesses behaviour
README.md - Hampel + STL → Adaptive Detection: supplies residuals
README.md - Adaptive Detection → Detection Filtering: detects candidates
README.md - Event Calendars → Metrics Evaluation: provides labels
README.md - Detection Filtering → Metrics Evaluation: supplies detections
README.md - Metrics Evaluation → Detected Events CSV: exports events
README.md - Station-Year Pipeline → Station Plots: plots single station
README.md - Detected Events CSV → Event Heatmap: reads detections
Python/heatmap.py