Skip to content

cran-comments.md

Latest commit

 

History

History
65 lines (41 loc) · 4.39 KB

cran-comments.md

File metadata and controls

65 lines (41 loc) · 4.39 KB

R CMD check results

0 errors | 0 warnings | 1 note

  • This is a new release. The note states, " Adding so many packages to the search path is excessive and importing selectively is preferable." Each package is used every time the function is run. Removing any package will reduce the accuracy of the results and/or the results returned to the user.

  • Added barchart to provide results of the Komogrov-Smirnoff test. The test determines if two samples came from the same population. The two samples tested are the predictions for each of the 40 models, and the true values. The null hypothesis for this test is the two samples are from the same popluation. For example, this will test if the Ensemble BayesRNN predictions, and the true holdout values come from the same population. Same for each of the 40 models. Two dotted blue lines are given to guide the user as a starting point, but the user may select any p-values they wish: p = 0.10 and p = 0.05. If the p-value for a model (such as SVM) is below the desired value, the test indicates there is sufficient evidence to reject the null hypothesis. Otherwise there is not sufficient evidence to reject the null hypothesis.

  • Added a small function to remove NAs from the ensemble. The ensemble virtually never has NAs, but this is just in case.

  • Please reduce the length of the title to less than 65 characters. Done

  • The Description field is intended to be a (one paragraph) description of what the package does and why it may be useful.

  • Please add more details about the package functionality and implemented methods in your Description text. Done, thank you for the note, the new description has more the motivation for the ensemble, and how it can be used to solve real problems

*Please add small executable examples in your Rd-files to illustrate the use of the exported function but also enable automatic testing. All my attempts to create an example that runs in five seconds or less have failed. This package totally automates the entire analytics process for numeric data, and takes more than five seconds even with the smallest data set.

  • You write information messages to the console that cannot be easily suppressed. Thank you for the note, changed all print commands to message.

  • Please ensure that your functions do not write by default or in your examples/vignettes/tests in the user's home filespace (including the package directory and getwd()).

  • This is not allowed by CRAN policies. Please omit any default path in writing functions. In your examples/vignettes/tests you can write to tempdir(). This took a very large amount of work, but is fixed. Nothing writes to the user's home filespace now, it all writes to tempdir1.

  • 0.9.2 updated the vignette so the code runs correctly, and corrected /R/NumericEnsembles.R code, so that set.seed runs correctly. (there was a line of code that was randomizing the data, that has been fixed).

  • 0.10.0, updated to show that NumericEnsembles runs 32 models now (used to run 40). The package runs in much less time than previously. Therefore the Vignette was updated to show the greatly improved speed.

  • 0.10.2, updated the ReadMe to reflect most recent changes, including adding a Quarto template for a comprehensive report on the data and analysis.

  • 0.10.3, update version number.

  • 0.10.4, remove all references to MAE, SSE, MSE (data, calculations, plots, etc.)

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

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

  • 0.10.7 Added stratified random sampling for the train, test and validation data sets

  • 0.10.8 Removed Ensemble_XGBoost due to massive overfitting, replaced with ensemble_neuralnet, which does not have any issues with overfitting.

  • 1.0.0, NumericEnsembles has all the desired functionality and features, all running very well. This is version 1.0.0.

  • 1.0.1 Corrected mistakes in the vignette.

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

  • 1.0.3 fixed errors in predicting on new data using XGBoost

  • 1.0.4, fixed error that the variable importance barchart was not being saved when the user requested it, and set_seed was crashing (due to the VIF function). Both are now fixed and working fine.