A Simple Fuzzy Classifier based on Inconsistency Analysis of Labeled Data

Obtaining accurate but also interpretable fuzzy rulebased classifiers from labelled observation data

An extremely simple fuzzy classifier is identified based on the inconsistency analysis of labelled training data. The method was applied to the COIL challenge 2000 Direct Mail problem and resulted in 121 selected caravan policies within the first 800 selected customers. As this result is identical to the result of the winner of the competition, the presented method is an example for how the try the simplest first approach can be effective in real-life problems.

This algorithm can be used for obtaining accurate but also interpretable fuzzy rule-based classifiers from labelled observation data. In the first step, the structure of the model is initialised based on the statistical analysis of the labelled data and straightforward data-mining tools like feature selection methods. After the feature selection step, the algorithm transforms the inconsistency analysis of the features into fuzzy sets. Finally, the classifier is optimised for accuracy by adapting the parameters of the resulting model.

The algorithm is also described in:
J. Abonyi, H. Roubos, Simple fuzzy classifier based on inconsistency analysis of labeled data", Chapter 12 in: CoIL Challenge 2000: The Insurance Company Case, Peter van der Putten and Maarten van Someren (eds), Sentient Machine Research, Amsterdam and Leiden Institute of Advanced Computer Science, Leiden, LIACS Technical Report, 1-10, 2000

More MATLAB implementation on my website:
http://www.abonyilab.com/software-and-data