Added Column-Based Windows (#151).
Refactored Data Slice Generator (#150).
Updated README (#164).
Updated Predict Next Purchase Demo (#154).
Updated Predict Turbofan Degradation Demo (#154).
Attributes of the data slice context have changed. Inside a labeling function, the timestamps of a data slice can be referenced by ds.context.slice_start and ds.context.slice_stop. For more details, see Data Slice Context.
Target values can be sampled from each group (#138).
One of multiple targets can be selected (#147).
Labels can be binned using infinite edges represented as string (#133).
The label times object was refactored to improve design and structure (#135).
Loading label times from previous versions will result in an error.
Label Search for Multiple Targets (#130)
Column renamed from cutoff_time to time (#139)
Dropped Support for Python 3.5 (#128)
Rename LabelTimes.name to LabelTimes.label_name (#126)
Support keyword arguments in Pandas methods. (#121)
Improved data download in Predict Next Purchase (#76)
Added tests that use Python 3.8 in CirlceCI (#128)
LabelTimes.name has been renamed to LabelTimes.label_name
Support for Pandas 1.0
Added higher-level mappings to offsets.
Track settings for sample transforms.
Pinned Pandas version.
Moved Featuretools to test requirements.
Serialization for Label Times
Matplotlib Backend Fix
Sampling Label Times
Added Data Slice Generator Guide
Integration Tests for Python Versions 3.6 and 3.7
Added Slice Generator
Added Seaborn Plots
Added Data Slice Context
Added Count per Group
Added Example: Predict Next Purchase
Added Example: Predict RUL
Added Sample Transform
Improved Progress Bar
Improved Label Times description
Added testing for Featuretools compatibility
Improved description of Label Times
Refactored search in Label Maker
Improved testing for Label Transforms
Add dynamic progress bar
Add label transform for binning labels
Improve code coverage