Skip to content

NEWS.md

Latest commit

 

History

History
116 lines (71 loc) · 4.83 KB

NEWS.md

File metadata and controls

116 lines (71 loc) · 4.83 KB
Raw

NumericEnsembles 1.2.1

Documentation & Compliance Updates

  • Added complete, CRAN-compliant roxygen2 @param tags to all primary exported pipeline functions to ensure complete .Rd manual generation.
  • Fully documented previously undocumented core user-level assets, including ExportNumericResults(), LaunchNumericApp(), NumericEnsemblesDemo(), RenderExecutiveReport(), load_pipeline(), predict_production(), and save_pipeline().
  • Enhanced DESCRIPTION field to satisfy CRAN's 3-sentence minimum structural rule and refined package dependency metadata.

Algorithmic & Performance Enhancements

  • Implemented defensive warning filters within the core engine's predictive evaluation sub-routine (suppressWarnings on ks.test) to prevent terminal logging pollution caused by dataset tie-points.
  • Standardized validation checks for input execution tracks, confirming parameter metrics adhere strictly to decimal fractions where proportional values are required.

Bug Fixes

  • Resolved missing argument matching warnings within Numeric.Rd and NumericEnsemblesConfig.Rd usage specifications.

NumericEnsembles 1.2

NumericEnsembles 1.1

Removed functions based on vif packacge because I received information the vif package will be archived

NumericEnsembles 1.0.5

1.0.4

Fixed errors that the variable importance barchart was not saving when the user requested it, and set_seed was crashing. Both are fixed, passed tests, running fine.

1.0.3

Fixed errors on predicting on new data using XGBoost

1.0.2

I received word that the package reactablefmtr will be archived on 2026-03-17. I replaced those functions in NumericEnsembles using the htmlwidges and htmltools libraries.

1.0.1

Corrected mistakes in the vignette.

1.0.0

All desired features are included and working great, ready for version 1.0.0!

0.10.10

Added stratified random sampling

Added extensive documentation, set up sections so the code is easy for the user to navigate.

0.10.9

Added several functions to separate best performing from worst performing data, using the vip package

0.10.8

Removed ensemble_XGBoost due to problems with overfitting, replaced with ensemble_neuralnet, which does not have any issues with overfitting.

0.10.7

Added automatic stratified sampling for train, test and validation sets. Also added automatic histograms of the overfitting data.

0.10.6

Add back remove_data_correlations_greater_than. This addresses issues when data has columns that are highly correlated.

0.10.5

Fix name: Change holdout / RMSE to overfitting (both more accurate and easier to understand)

0.10.4

Remove all references to MAE, MSE, SSE

0.10.3

Update version number to reflect all changes.

NumericEnsembles 0.10.2

Updated ReadMe to reflect recent changes (32 models instead of 40), added a report (in the Vignettes) in Quarto format

NumericEnsembles 0.10.1

0.9.2

Corrected errors in the vignette, which now runs correctly

Corrected errors in set.seed, so the results are consistent if the same seed is used.

NumericEnsembles 0.9.0

Corrected 160 plots (40 models x 4 plots each, pred_vs_actual, pred_vs_residuals, hist_residuals and qq plots. They were not printing the correct plot, that is now fixed.)

Added set_seed, so NumericEnsembles runs correctly when a specific seed is set.

NumericEnsembles 0.8.0

Added Variance Inflation Factor. The user is able to set the VIF value, and models are built with VIF values at or below the user's choice of VIF value

Added "free" and "fixed" scales to all appropriate plots. Each result has two plots, one with free scales, the other with fixed scales.

Added Kolomogrov-Smirnovv test to help the user see which models test similar to the actual holdout data

Added several "Holdout vs train" charts to show how each model performs across multiple resamples, and the range of values of holdout RMSE / train RMSE

NumericEnsembles 0.7.0

NumericEnsembles 0.6.0

NumericEnsembles 0.5.0

NumericEnsembles 0.4.0

NumericEnsembles 0.3.0

NumericEnsembles 0.2.0

NumericEnsembles 0.1.0

  • Initial CRAN submission.

NumericEnsembles 1.0.0

Added example of New_Boston data set as a new data set to use in NumericEnsembles ('do_you_have_new_data?')

  • To see how this works, use the data as Boston_housing, and the new data as New_Boston.

Removed neuralnet models since I could not get the RMSE down, added cppls, which creates a much more accurate ensemble compared to the previous version.

Removed cppls, not reliable on several data sets.

Re-added neuralnet models (individual and ensemble) since I was able to get those to work without error

Added best subsets: Forward, backward, exhaustive and seqrep

Added save_all_plots to automatically save all plots in the user's choice of one of six graphics formats: eps, jpeg, pdf, png, svg or tiff