How knamlx/ionoautoml works
IonoAutoML is a research pipeline for collecting ionospheric and geophysical observations, preparing station time series, engineering forecasting features, and evaluating foF2 models. The main workflow runs from external data sources through cleaning, time-grid normalization, optional filtering, feature creation, and station-wise ML experiments; outputs support quality and latitude-zone analysis. The README documents the stages and capabilities, while sampled source confirms collection, transformation, filtering, feature generation, model fitting, and batch dispatch. Some subsystem internals are unsampled, so only documented or directly evidenced links are shown.
Main entry points: collect_hf_data.py Data collection · normalize_time_grid.py Time normalization · run_ml_baselines.py Model training · clean_collected_data.py Data cleaning
Components 24
Data collection
- Data collection
collect_hf_data.py - Raw observations
Data preparation
- Data cleaning
clean_collected_data.py - Cleaned tables
- Time normalization
normalize_time_grid.py - Station time grids
- Series filtering
filter_time_series.py - Feature engineering
build_features.py - Feature datasets
- Station selection
configs/station_selection.json
Forecasting
- Experiment settings
configs/experiments - Station batch runner
run_station_batch.py - Model training
run_ml_baselines.py - Walk-forward evaluation
run_ml_baselines.py - Metrics and predictions
- Run progress
watch_run_progress.py
Results analysis
- ML results review
notebooks/ml_results_review.ipynb - Latitude-zone analysis
scripts/build_latitude_zone_article.py - Analysis reports
reports
Other components
- Researcher
- GIRO DIDBase
- GFZ Kp API
- NASA OMNIWeb
- NOAA SWPC
Connections 28
- Researcher → Data collection: starts collection
README.md - Data collection → GIRO DIDBase: fetches observations
README.md - Data collection → GFZ Kp API: fetches indices
README.md - Data collection → NASA OMNIWeb: fetches solar wind
README.md - Data collection → NOAA SWPC: fetches live data
README.md - Data collection → Raw observations: writes raw data
README.md - Researcher → Data cleaning: runs cleaning
README.md - Raw observations → Data cleaning: reads tables
README.md - Data cleaning → Cleaned tables: writes cleaned tables
clean_collected_data.py - Researcher → Time normalization: runs normalization
README.md - Cleaned tables → Time normalization: reads processed data
README.md - Time normalization → Station time grids: writes station grids
README.md - Researcher → Series filtering: runs filtering
README.md - Station time grids → Series filtering: reads time grids
README.md - Series filtering → Feature engineering: filtered input
README.md - Station time grids → Feature engineering: reads station data
README.md - Feature engineering → Feature datasets: writes training tables
build_features.py - Station selection → Time normalization: configures selection
README.md - Researcher → Station batch runner: starts station runs
README.md - Station batch runner → Model training: dispatches station jobs
run_station_batch.py - Experiment settings → Model training: configures experiment
README.md - Feature datasets → Model training: provides features
README.md - Model training → Walk-forward evaluation: evaluates forecasts
run_ml_baselines.py - Walk-forward evaluation → Metrics and predictions: saves metrics
README.md - Metrics and predictions → Run progress: reads run status
watch_run_progress.py - Metrics and predictions → ML results review: reviews results
README.md - Metrics and predictions → Latitude-zone analysis: summarizes model quality
scripts/build_latitude_zone_article.py - Latitude-zone analysis → Analysis reports: builds analysis reports
README.md