Python analytical scripts that will overcome paralysis in your data analysis.
What is it?
Paralytics package was created in order to simplify and accelerate repetitive tasks during modeling and predictive analysis. It especially puts stronger emphasis on data preprocessing, which is often the most arduous stage of modeling.
The purpose of this package is to reduce to a minimum time allocated on repetitive activities preceding the problem-specific approach to a given problem, containing among others optimization of the applied machine learning techniques, which is the part that most of Data Scientists would like to devote the most energy to, however, by poorly prepared data, it is often only a fraction of the total work time devoted to the project.
Highlighting the main functionalities of the Paralytics:
Expanded target encoding of categorical variables using double cross-validation technique with additional regularisation preventing favoritism of sparse categories with reduction of excessive adjustment to the training set, effectively reducing overfitting.
Discretization of continuous variables to ordinal using shallow decision tree or method based on Spearman’s rank-order correlation.
Processing data read into the DataFrames, including:
automatic unification of variable types,
grouping of sparse categories,
projecting of text variables whose unique elements symbolize a binary response onto binary variables,
imputation of missing data.
Collinearity reduction using such factors as: variance inflation factor (VIF) or correlation.
Paralytics package requirements are checked and, if needed, installed during the installation process automatically. Mainly used packages across the Paralytics are:
NumPy (>= 1.15.4)
Pandas (>= 0.23.4)
scikit-learn (>= 0.20.1)
SciPy (>= 1.1.0)
matplotlib (>= 3.0.2)
The easiest way to install the package is using pip:
pip install paralytics
If you want to omit the packages upgrade (when your version does not meet the requirements) run:
pip install paralytics --no-deps
When functionalities requiring optional dependencies are needed you can install those extra requirements by running:
pip install paralytics[<extra-name>]
For example, to use BaseSeleniumBrowser (requires selenium) and VIFSelector (requires statsmodels) run:
pip install paralytics[browser,vif]
Installation of all extras is possible via:
pip install paralytics[all]
You can always install directly from the github repository:
pip install git+https://github.com/mrtovsky/Paralytics.git
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.