Add wildfire baseline forecasting workflow for AI013

[thewonderworking]

Summary

This PR adds my AI013 wildfire baseline forecasting contribution inside the time-series model folder.

Work Completed

• Added a wildfire baseline forecasting notebook under ai-ml/models/time-series
• Used acq_date as the timestamp
• Used region as the location field
• Used frp_mw as the target variable for daily fire intensity
• Filtered the wildfire data for Australia
• Aggregated wildfire observations into a daily time series
• Created a continuous daily date range
• Created calendar, lag, and rolling average features
• Tested three baseline forecasting models:
  • Naive forecast
  • Seven-day moving average forecast
  • Linear Regression with lag features
• Evaluated the models using MAE, RMSE, and MAPE

[SunainM]

W work, but mostly irrelevant to Project.

[SunainM]

This is useful standalone baseline work, but it should not be merged yet as-is.

Required fixes:

1. Add the dataset path properly or document exactly where wildfire_multi_region_dataset.csv comes from. (use universal naming)
2. Share any locally used csv/dataset with team
3. Add a short README inside ai-ml/models/time-series/abin-wildfire-baseline/ explaining:
   • dataset source
   • target variable
   • how to run
   • expected outputs
   • model purpose
4. Connect the notebook back to AI013/ADCRS by explaining how the forecast output can support risk scoring.
5. Preferably convert the reusable parts into a small Python module or pipeline-compatible script instead of notebook-only code.
6. If this is only exploratory work, label it clearly as baseline_experiment and not final AI013 integration.

Good work technically, but currently it is not integrated enough with the PHOENIX AI/ML pipeline or ADCRS purpose.