Skip to content

Latest commit

 

History

History
123 lines (86 loc) · 5.7 KB

NEWS.md

File metadata and controls

123 lines (86 loc) · 5.7 KB
Raw

SDMtune 1.3.1

  • Minor improvements in documentation (#27)

SDMtune 1.3.0

  • Add function to combine cross validation models
  • Improve show method for package outputs
  • Fix CRAN note for undeclared package in Rd file

SDMtune 1.2.1

  • plotCor()gains the argument text_size to tweak the size of the text in the plot
  • Cross validation folds can be created with blockCV version 3.x (#24)
  • Support for the raster package has been deprecated (#23)

SDMtune 1.2.0

  • SDMtune migrated to package terra to handle spatial data. Objects created with the package raster will not be supported in future releases
  • SDMtune has a new progress bar and messages/warnings/errors have a new look
  • modelReport(), prepareSWD, reduceVariable(), thinData(), and varSel() gain the argument verbose to disable informative messages
  • gridSearch(), optimizeModel(), randomSearch(), reduceVar(), and varSel() gain the argument interactive to disable the interactive real-time chart
  • doJk(), gridSearch(), optimizeModel(), predict(), randomSearch(), thinData(), train(), varImp(), and varSel() gain the argument progress to disable the progress bar
  • The function thinData() prints a message with the number of removed locations
  • The argument factors in modelReport() is deprecated. See terra documentation to see how to use factors during predictions

Bug fixes:

  • Real-time charts are again displayed in Windows
  • plotCor() plots a correct heatmap when there are more than 9 variables
  • doJk() runs also without testing dataset

SDMtune 1.1.6

Resolve CRAN notes

SDMtune 1.1.5

Bug fixes:

  • Allow fold partitions created with ENMeval >= 2.0
  • Fix error generated when plotting response for categorical predictors using RF
  • Update maxnet to version 0.1.4 to solve issue in cran checks

SDMtune 1.1.4

Bug fixes: Add the argument factors to the function modelReport() to allow predictions for raster objects including categorical variables

SDMtune 1.1.3

Include Java >=8 in system requirements

SDMtune 1.1.2

  • Removed deprecated function get_tunable_args().
  • Removed deprecated argument "parallel" form functions.
  • Updated citation text

SDMtune 1.1.1

Main changes:

  • New function checkMaxentInstallation() to check if Maxent is correctly configured.
  • The argument parallel is deprecated and not used anymore. This because it improved computation only for very large datasets and used the superseded package snow.
  • The function get_tunable_args() has been replaced by getTunableArgs() to be consistent with the camel case function naming. At the moment is still possible to use both functions but get_tunable_args() will be deprecated in the next release.
  • Four new vignette are now available with the package.

Bug fix: plotPA function now works also with new version of ggplot2.

SDMtune 1.1.0

Main changes:

  • The function thinData accept now a matrix or a dataframe with several columns, useful if the users has information related to the coordinates that doesn't want to lose with the thinning procedure.
  • The function plotResponse plots the response for the full range of presences and backgrounds/absences when only_presence is TRUE, only_presence is unused only to compute the provided function to the range of presence locations when marginal = TRUE

Bug fix:

  • Interactive plot of SDMtune objects are again displayed in the RStudio Viewer pane

SDMtune 1.0.1

Bug fixes:

  • Fix CRAN errors
  • Fix bug introduced with version 1.0.0

SDMtune 1.0.0

Main changes:

  • In this release all deprecated functions and functions' arguments have been removed, including the functions to convert old objects (created with version < 0.2.0.) into the new format.
  • Add function addSamplesToBg to add presence locations to background location.
  • extra_args in Maxent models cannot be changed anymore.

Bug fix: Fix CRAN error for r-oldrel

SDMtune 0.2.1

Main changes:

  • New function to predict SDMmodelCV objects
  • Add the possibility to pass multiple methods to the train function
  • Add possibility to compute the testing AUC or TSS for a held apart testing dataset in the case of a SDMmodelCV object
  • Add the possibility to merge only the presence locations in the mergeSWD function

Bug fixes:

  • plotPA function now works also with large raster objects
  • Removed font family from plot functions to avoid errors when the font is not available

SDMtune 0.2.0

This release is the first step to enable more methods to train models. The main change is that the SDW object now bundles together the presence and the absence/background locations and is not necessary anymore to pass the presence and absence locations as separate arguments to the train function.

Old objects of class SWD, SDMmodel, SDMmodelCV and SDMtune created with version <= 0.1.1 must be converted into the new format using the dedicated help functions.

Main changes:

  • The SWD object bundles together the presence and absence/background locations
  • New methods to train models: Artificial Neural Network (ANN), Boosted Regression Trees (BRT) and Random Forest (RF)
  • Enabled spatial cross validation: the function train accepts now folds partition generated with other packages (i.e. ENMeval and blockCV)
  • Add title argument in plot function for SDMtune objects
  • New randomly generated dataset virtualSp with presence, absence and background locations
  • Number of background locations cannot be tuned anymore using the tuning functions

Deprecated function: getSubsample

SDMtune 0.1.1

  • Fix bug in VarImp function for SDMmodelCV objects
  • Real-time charts are now displayed also when R is not executed from RStudio
  • maxentVarImp function available for SDMmodelCV objects

SDMtune 0.1.0

  • First release