Data-driven fuzzification technique (including tutorial)

This submission is somewhat the software implementation of the technique proposed in https://www.tandfonline.com/eprint/z2MHECZWBZFQw72K5vwp/full (the link provides the full access to the paper). Besides implementing the aforementioned technique, the submitted tool provides some allegedly useful additional functionality that is described further.

The submission is partially devoted to the fuzzification issue in linguistic summarization (LS) of data. LS is a branch of Data Mining with goal to extract [from data] knowledge elements being in linguistic form. This submission deals with the Wu–Mendel approach for linguistic summarization of datasets. This approach and the submitted tool can interest someone for the following reasons:
1) One can get linguistic descriptions of causal datasets (datasets with antecedent and consequent attributes).
2) One can get a rulebase for some Fuzzy Inference System (FIS) by applying the Wu–Mendel approach. This means that the Wu–Mendel approach can be used to design fuzzy prediction systems from statistical data: one gets a causal dataset (for example, from UCI Machine Learning Repository) and, by using the submitted tool, creates a Mamdani-Type FIS that predicts the values of the dataset's consequent attribute/attributes.
3) The submitted tool implements the fuzzification-tuning technique, which can be quite useful in both obtaining linguistic descriptions (3.1) and obtaining the rulebase for a FIS (3.2).
3.1) The attributes of a dataset to be put through LS are usually fuzzified: each attribute gets a number of fuzzy linguistic terms (for example, an attribute "Income" can get the following fuzzy linguistic terms: Low, Middle, High). The fuzzification is usually user defined in LS; so the current submission implements the technique that tunes the user-defined fuzzification by merging highly cognate [and therefore redundant] linguistic terms.
3.2) The fuzzification-tuning technique can be used to decrease the number of membership functions and rules in FIS being created by means of the Wu–Mendel approach without the proportionately siginificant loss in inference accuracy.
4) Theoretically, the fuzzification-tuning technique can be used as a common multipurpose fuzzification-tuning technique in Data Mining.