DI2 1.0.2

An an unsupervised discretization method, DI2, for variables with arbitrarily skewed distributions.

DI2 (Distribution Discretizer)

Discretizer

distribution_discretizer(dataset, number_of_bins, statistical_test, cutoff_margin, kolmogorov_opt, normalizer, distributions, single_column_discretization)

distribution_discretizer(pandas.Dataframe, integer, optional:string, optional:float, optional:boolean, optional:string, optional:array) - Discretizes data according to the best fitting distribution.

The distribution_discretizer(pandas.Dataframe, integer, string, float, boolean, string, array) receives the data (pandas.Dataframe), an integer representing the number of categories for discretization, an string with the name of the main statistical hypothesis test to apply (options available: "chi2", "ks"), a float between 0 and 0.49 which indicates the width range to consider a value as being a border value, a boolean indicating if outliers should be removed, a string indicating the normalization method to be used (options available: "min_max","mean", z_score), an array of continuous distributions, from https://docs.scipy.org/doc/scipy/reference/stats.html, to be considered by the discretizer.

Data normalization:

Normalizes data by min_max normalization https://en.wikipedia.org/wiki/Feature_scaling#Rescaling

Normalizes data z-score normalization https://en.wikipedia.org/wiki/Feature_scaling#Standardization

Normalizes data by mean normalization https://en.wikipedia.org/wiki/Feature_scaling#Mean_normalization

Goodness of fit test

Person's chi-squared goodness of fit test https://en.wikipedia.org/wiki/Pearson%27s_chi-squared_test

Kolmogorov–Smirnov goodness of fit test https://en.wikipedia.org/wiki/Kolmogorov%E2%80%93Smirnov_test

As an illustrative example we use the dataset available at the UCI machine learning repository https://archive.ics.uci.edu/ml/datasets/Breast+Tissue.

---> DI2 was developed by L. Alexandre ([e-mail]), R.S. Costa ([e-mail]) and R. Henriques <---