What's New in VerticaPy v0.6.0

New features:

Machine Learning

• AutoDataPrep is now available. AutoDataPrep automatically finds relations between the different features to preprocess the data according to each column type.
• Improved AutoML is now available. AutoML is a powerful machine learning technique that automatically tests several models and returns the one with the highest input score.
• AutoClustering is now available. AutoClustering creates k groups and for generalizing the data.
• New method: Model.contour draws the contour plot of your model.
• New method: Model.to_sql exports your model as regular SQL.
• New method: Model.to_python exports your model as an independent Python function.
• New functions for evaluating models and tuning hyperparameters: bayesian_search_cv, enet_search_cv, randomized_features_search_cv and stepwise.

Geospatial

• New Functions: coordinate_converter converts bertween latitude and longitude and Euclidean coordinates (x,y) and to split any polygon to n identical smaller ones.
• New Functions: split_polygon_n splits any polygon to n identical smaller polygons.

vDataFrame

• New method: vDataFrame[].cut allows you to discretize numerical features.
• New method: vDataFrame.add_duplicates allows you to add duplicates when it is needed. An example could be creating a weighted KMeans.
• Additional parameter 'ncols_block' for the vDataFrame.describe and vDataFrame.agg methods: this parameter lets you control the number of columns specified in the generated SQL code and can make queries "lighter" according to your use case.

Graphics and Data Visualization

• A loading bar is available for many functions. It uses 'tqdm' API.
• New method: vDataFrame.animated draws animated charts.
• New method: vDataFrame.contour draws contour plots.

Datasets

• New time series datasets: load_gapminder and load_pop_growth.
• You can now generate your own datasets with gen_dataset and gen_meshgrid.

Deprecated, Replaced, or Moved

• The package verticapy.connections.connect has been moved to verticapy.connect
• The package verticapy.learn.datasets has been moved to verticapy.datasets
• The package verticapy.learn.tsa.tools has been moved to verticapy.stats
• The package verticapy.learn.tsa.models has been moved to verticapy.learn.tsa
• The functions drop_table, drop_text_index, drop_view, and drop_model have been replaced by the drop function.
• The function load_model has been moved from utilities to learn.tools.

GitHub and Unit Tests

• More unit tests are available.