What’s new¶
v0.6.0¶
Enhancement¶
- Support pandas v0.22.0 and scikit-learn 0.20.0.
API Change¶
- ModelFrame.model_selection.describe now returns ModelFrame compat with GridSearchCV.cv_results_
Deprecation¶
- Drop support of pandas v0.18.x or earlier
- Drop support of scikit-learn v0.18.x or earlier.
v0.4.0¶
Enhancement¶
- Support scikit-learn v0.17.x and v0.18.0.
- Support imbalanced-learn via
.imbalance
accessor. See Handling imbalanced data. - Added
pandas_ml.ConfusionMatrix
class for easier classification results evaluation. See Confusion matrix.
Bug Fix¶
ModelFrame.columns
may not be preserved via.transform
usingFunctionTransformer
,KernelCenterer
,MaxAbsScaler
andRobustScaler
.
v0.3.1¶
Enhancement¶
inverse_transform
now reverts originalModelFrame.columns
information.
Bug Fix¶
- Assigning
Series
toModelFrame.data
property raisesTypeError
v0.2.0¶
Enhancement¶
ModelFrame.transform
can preserve column names for somesklearn.preprocessing
transformation.- Added
ModelSeries.fit
,transform
,fit_transform
andinverse_transform
for preprocessing purpose. ModelFrame
can be initialized fromstatsmodels
datasets.ModelFrame.cross_validation.iterate
andModelFrame.cross_validation.train_test_split
now keep index of original dataset, and addedreset_index
keyword to control this behaviour.
Bug Fix¶
target
kw may be ignored when initializingModelFrame
withnp.ndarray
andcolumns
kwds.linear_model.enet_path
doesn’t accept additional keywords.- Initializing
ModelFrame
with namedSeries
may have duplicated target columns. ModelFrame.target_name
may not be preserved when sliced.
v0.1.0¶
- Initial Release