hungnk25 / dython

Python Data Tools

Dython

A set of Data analysis tools in pYTHON 3.x.

Installation:

Clone this repository to your local machine anr run pip:

git clone https://github.com/shakedzy/dython.git
cd dython
pip install -e .

Dependencies: numpy, pandas, seaborn, scipy, matplotlib, sklearn

Nominal tools (nominal.py):

A set of functions to explore nominal (categorical) datasets and mixed (nominal and continuous) data-sets.

Coefficients and statistics:

• Conditional entropy (conditional_entropy)
• Cramer's V (cramers_v)
• Theil's U (theils_u)
• Correlation ratio (correlation_ratio)

Additional functions:

• associations: Calculate correlation/strength-of-association of a data-set
• numerical_encoding: Encode a mixed data-set to a numerical data-set (one-hot encoding)

Model utilities (model_utils.py)

A set of functions to gain more information over a model's performance.

• roc_graph: compute and plot a ROC graph (and AUC score) for a model's predictions
• random_forest_feature_importance: plot the feature importance of a trained sklearn RandomForestClassifier
• associations: Calculate correlation/strength-of-association of a data-set (same as nominal.associations)

Examples:

See the examples.py module for roc_graph and associations examples.

License:

Apache License 2.0