OneR 2.2 (2017-05-05)
=====================

MAJOR IMPORVEMENTS
- OneR: massive speedup of main OneR function (> 30 times faster).
- OneR & optbin: standard S3 method interface added for formulas and data frames.

MINOR IMPROVEMENTS
- optbin: speedup of method "infogain".
- Some minor corrections in documentation.


OneR 2.1 (2016-10-24)
=====================

NEW FEATURES
- eval_model: all instances of "prediction" and "actual" are now being printed in the confusion matrices. Two new arguments were added: "dimnames" for the printed dimnames of the confusion matrices and "zero.print" specifying how zeros should be printed; for sparse confusion matrices, using "." can produce more readable results. A new performance measure "error rate reduction versus the base rate accuracy" was added together with a p-value.

MINOR IMPROVEMENTS
- Some minor corrections in documentation.
- Some streamlining and consolidation of code for better maintenance.


OneR 2.0 (2016-08-12)
=====================

NEW FEATURES
- Added a vignette.
- breastcancer: Breast Cancer Wisconsin Original Data Set now included in the package.
- predict: new type "prob" which gives a matrix whose columns are the probability of the first, second, etc. class.
- optbin: new method "infogain" (information gain) which is an entropy based method to determine the cutpoints which make the resulting bins as pure as possible.
- OneR, optbin, maxlevels: consistent handling of unused factor levels (e.g. due to subsetting) was added. These are dropped for analysis and a warning is given.

MINOR IMPROVEMENTS
- bin & optbin: in case of removing instances due to missing values the resulting warning gives the number of removed instances.
- maxlevels: with data containing missing values an unhelpful warning was given.
- predict: numerical values that are smaller or bigger than model limits are now transformed into (-Inf, min] or (max, Inf] respectively.
- predict: output of type "class" is a factor now.
- Some streamlining and consolidation of code for better maintenance.

BUGFIXES
- bin & optbin: in some borderline cases when the function addNA was used in preprocessing print.OneR stopped with an error.


OneR 1.3 (2016-07-22)
=====================

NEW FEATURES
- bin: new method "clusters", which determines the bins according to automatically determined clusters in the data.
- OneR: a new element "call" with the specified arguments of the actual function call was added to the internal class structure of OneR objects.
- print & summary: the function call with the specified arguments which was used to build the model is printed first.

MINOR IMPROVEMENTS
- bin & optbin: in cases where there were missing values and already a factor level "NA" the functions gave an unhelpful warning.
- eval_model: added warning when actual contains missing values.
- eval_model: added "Confusion matrix" to printout for clarity.
- Extension of and minor corrections in documentation
- Some minor streamlining of code.

BUGFIXES
- predict: the combination of intervals and "NA"s caused an error.
- bin: the method "content" stopped with an error in case of missing values.
- optbin: the method "logreg" stopped in some borderline cases with missing values with an error.
- optbin: some borderline cases could result in a "breaks are not unique" error.
- OneR: in some borderline cases with very large datasets the numbering of printed ranks (verbose = TRUE) could be wrong due to rounding errors.


OneR 1.2 (2016-06-20)
=====================

Initial release on CRAN
