Final update before resubmitting to CRAN and to JOSS
Vignettes were removed from CRAN release considering the new NOTE raised and the absolute lack of insights on what could cause it.
Slight modification in the vignettes and documentation. Adding an error when the user gives data using the old packages raster and sp.
Correcting minor bugs caused by the recent removing of rgdal from dependencies Updating to C++17 to match CRAN new requirements
Removing dependencies from rgdal and raster to move to terra. The dataset Aracachon is now provided as a raw tiff file and must be read directly.
Adding two new parameters to the functions CMeans, GCMeans, SFCMeans, SFGCMeans : noise_cluster and robust. They can be used to calculate the “robust” version of each algorithm, or to add a noise cluster. See the vignette “Advanced examples” on the website.
Adding a little function to facilitate the scaling and unscaling of variables : standardizer
Replacing all the functions from maptools, sp and rgeos to work now with feature collections from sf.
removed the old function future::multiprocess, for future::multisession as suggested in issue #3
Corrected the bug in the issue #2
Minor release for correcting minor bugs and providing an updated documentation.
New Features
- Added support to use raster data for clustering (see vignette rasters)
- Added a S3 method to predict the membership matrix of a new set of observations (predict.FCMres)
- Added a shiny app (function: sp_clust_explorer) for result exploration
- The results of the functions CMeans, GCMeans, SFCMeans, SGFCMeans are now objects of class FCMres and the generic methods predict, summary, plot, is and print can be used on them. FMCres object can easily be created by hand with results from other classifier if needed, see the new vignette FMCres.
- Added some clustering quality indices : Negentropy Increment index, Generalized Dunn’s index (43 and 53), David-Bouldin index, Calinski-Harabasz index
- Added a function to perform clustering validation by bootstrap (see function bstp_group_validation)
- Added a function to reorder the results of a classification to match the most similar groups in a second classification (groups_matching)
- Added functions to evaluate spatial autocorrelation of a classification results: ELSA and FuzzyELSA (see functions calcELSA and calcFuzzyELSA and the end of the vignette rasters)
corrected bugs
- issue 1 fixed by editing the mapping functions. A bug occurred when the fid of a SpatialDataFrame read from a shapefile was different from 1:nrow(df)
- an important performance gain can be observed for large dataset, the function to compare matrices between two iterations is now significantly faster.
- core functions rewritten with Rcpp for massive time gain