User can now pass dataframes with inclusion of non-index-range column as index or with multi column index (requires named index columns). Such index types are added to the returned "ID" sets which are consistently shuffled and partitioned as the train and test sets.
2.16
1) User can now pass ML_cmnd to automunge function with 'PCA_cmnd':{'bool_PCA_excl':True} Which excludes returned boolean columns from any PCA dimensionality reduction. This may result in memory performance improvements for large datasets.
2) Corrected bug for labels processing in postmunge function
3) Corrected derivation of LabelFrequencyLevelizer, number of appended sets reduced by 1
4) Corrected bug for LabelFrequencyLevelizer of "singlect" (binary) category labels
2.15
added comparable treatment from version 2.14 updates for labels set to both validation labels sets returned from automunge(). (Move fast and fix things, that's our motto.)
2.14
apply ravel to returned labels numpy array if appropriate - converts e.g. [[1,2,3]] to [1,2,3]
2.13
a few clarifications associated with the new functionality for passing lists of ID column strings such as to trainID_column from version 2.13
2.12
corrected family tree derivations for primitives with offspring