Changelog

Source: NEWS.md

deweather 1.0.0

Major Version Changes

Version 1.0.0 of deweather is a complete re-write of the deweather package. This new version:

• Uses the tidymodels framework, allowing for more flexibility in plotting engines. deweather 1.0.0 launches with both xgboost and lightgbm engines available.

• Provides much more flexible partial dependency calculations, including grouped PDs.

• Uses the flexible mirai package to support parallelisation.

• Uses a consistent function and object naming scheme for easier applications.

The main reason prompting this change was the retirement of the gbm R package, and the slow development of gbm3. xgboost and lightgbm are modern, fast, popular, and highly capable implementations of gradient boosted machine learning.

The original version of deweather (including its NEWS.md) is archived at https://github.com/openair-project/deweather-archive for users still interested in the old API.

New Features

• Model building functions:

  • build_dw_model() fits a deweather model, used in the rest of the package.

  • tune_dw_model() allows for different modelling parameters to be tweaked and experimented with.

  • append_dw_vars() attaches a variety of modelling variables, and is used automatically within the above two functions.

  • The get_dw_pollutant() family allows for specific features of deweather models to be extracted consistently.

• Visualisation functions:

  • plot_dw_importance() provides a quick plot of variable importance of a deweather model.

  • plot_dw_partial_1d() calculates and visualises partial dependencies of any subset of model variables.

  • plot_dw_partial_2d() calculates and visualises two-dimensional partial dependencies.

• Modelling functions:

  • predict_dw() allows for the use of a deweather model for predictions.

  • simualte_dw_met() will simulate a timeseries in which selected meteorological variables are averaged, effectively helping ‘remove’ the influence of met variables.