AutoML toolbox - PROJECT DEPRECATED

The Automatic Machine Learning (AutoML) toolbox is a collection of methods with a simple API which can assist you in your Data Science tasks by providing boilerplate code in a form of simple functions ready to be used. You can use it as your personal Data Science assistant / wizard for creating ETL processes and/or a fully automated machine learning pipelines in a simple and quick way.

This library is intended to be used as a testing playground for a bunch of wrapper methods to serve as high-level APIs to a bunch of common tasks like:

• data profiling
• cleaning missing values
• detecting outliers
• performing feature engineering
• hyper-parameter optimization
• evaluating machine learning models
• creating ensembles of such models
• etc.

Warning

This code base is in heavy development for now. Once it reaches v0.1.0 you may then try it, but for now you are at your own risk.

Installation

For now, to install this package you must build it from source. To do that, just run the following command in the terminal:

python setup.py install

Note: once this package reaches v0.1.0 it will be possible to install it via pip.

Key Libraries used

This toolbox integrates the following packages in its core for doing most of its work. Basically, you can think of this package as a wrapper for a bunch functions you would uneed like cross-validation, hyperparameter optimization, etc., but with a nice, high-level API.

• Numpy
• Pandas
• pandas-profiling (data profiler)
• Scikit-learn (collection of ML libs)
• xgboost (ML lib)
• lightgbm (ML lib)
• Hyperopt (hyperparam optim - bo)
• HpBandSter (hyperparam optim - hyperband + bo)

Libraries to be integrated in the future

• dask (distributed computing / big data)
• keras (DL lib)
• feature-tools (automatic feature engineering)
• pygdf (GPU DataFrame)

TODO

Funcionalities intended to be added to the toolbox:

• basic data profiler
• automatic analysis / benchmarking and filling of missing values
• automatic analysis / benchmarking and cleaning of outliers
• automatic feature transformations / normalization
• automatic feature engineering
• automatic feature selection
• automatic model selection
• automatic model optimization (hyper-parameter optimization)
• automatic model ensembling
• pre-defined parameter list of the most popular ML models in scikit-learn
• distributed computing (integrate Dask)
• pipeline generation

License

MIT