How calvin4370/nhb-museum-visitorship-forecasting works
This repository forecasts monthly visitorship for five Singapore museums, fetching visitor and international-arrival series before preparing features, evaluating models, ranking them, and producing forecasts up to 24 months ahead. The README also documents saved-model regeneration, tuning reports, permutation importance, financial-year summaries, plots, and an IHC Deepavali analysis. The graph follows the documented and excerpted orchestration; connections to unsampled implementations are included only where the SOURCE INDEX or README supports them.
Main entry points: pipeline.py Museum pipeline · model_registry.py Model registry · main.py CLI entry · singstat_api.py SingStat client
Components 31
Data and features
- SingStat client
src/data/singstat_api.py - Raw series
data/raw - Feature engineering
src/features/data_prep.py - Event ranges
src/events/event_range.py - Prepared data
data/processed
Forecasting
- Model registry
src/models/model_registry.py - Baseline model
src/models/baseline.py - Holt-Winters
src/models/holtwinters.py - SARIMAX
src/models/sarimax.py - Random Forest
src/models/randomforest.py - XGBoost
src/models/xgboost_model.py - Support vector
src/models/svr.py - LSTM
src/models/lstm.py - TimeGPT adapter
src/models/timegpt.py - Fitted model wrapper
src/models/fitted.py
Analysis and reporting
- Optuna persistence
src/analysis/tuning.py - Model artifacts
src/analysis/artifacts.py - Permutation importance
src/analysis/importance.py - Importance report
src/visualisation/feature_report.py - Forecast plots
src/visualisation/timeplot.py - Seaborn plots
src/visualisation/jj_new_plots.py - Museum summaries
src/visualisation/summary.py - Deepavali analysis
src/analysis/deepavali.py - Tuning report
src/analysis/tuning_report.py - Forecast outputs
outputs
Runtime orchestration
- CLI entry
main.py - Argument selection
src/cli.py - Forecast configuration
config.py - Museum pipeline
src/pipeline.py
Other components
- Operator
- SingStat API
Connections 33
- Operator → CLI entry: runs
README.md - CLI entry → Argument selection: parses options
main.py - CLI entry → SingStat client: fetches series
main.py - CLI entry → Museum pipeline: runs museums
main.py - CLI entry → Forecast configuration: selects defaults
main.py - SingStat client → SingStat API: requests data
src/data/singstat_api.py - SingStat client → Raw series: returns series
README.md - Museum pipeline → Feature engineering: engineers features
src/pipeline.py - Feature engineering → Event ranges: loads occurrences
src/features/data_prep.py - Museum pipeline → Prepared data: saves features
README.md - Museum pipeline → Model registry: selects models
src/pipeline.py - Model registry → Baseline model: dispatches
src/models/model_registry.py - Model registry → Holt-Winters: dispatches
src/models/model_registry.py - Model registry → SARIMAX: dispatches
src/models/model_registry.py - Model registry → Random Forest: dispatches
src/models/model_registry.py - Model registry → XGBoost: dispatches
src/models/model_registry.py - Model registry → Support vector: dispatches
src/models/model_registry.py - Model registry → LSTM: dispatches
src/models/model_registry.py - Model registry → TimeGPT adapter: dispatches
src/models/model_registry.py - SARIMAX → Optuna persistence: creates study
src/models/sarimax.py - Holt-Winters → Optuna persistence: creates study
src/models/holtwinters.py - Museum pipeline → Model artifacts: saves models
src/pipeline.py - Museum pipeline → Permutation importance: computes importance
src/pipeline.py - Museum pipeline → Importance report: builds report
src/pipeline.py - Museum pipeline → Forecast plots: renders plots
src/pipeline.py - Museum pipeline → Seaborn plots: renders plots
src/pipeline.py - Museum pipeline → Museum summaries: writes summaries
src/pipeline.py - Museum pipeline → Deepavali analysis: builds analysis
src/pipeline.py - Museum pipeline → Tuning report: writes report
src/pipeline.py - Museum pipeline → Forecast outputs: writes forecasts
README.md - Fitted model wrapper → Model artifacts: saved by
- Tuning report → Optuna persistence: loads studies
src/analysis/tuning_report.py - Museum summaries → Deepavali analysis: includes table
src/visualisation/summary.py