Evaluation of the quality and relevance of a fuzzy partition

This paper proposes a couple of criteria for evaluating the quality and relevance of a fuzzy partition. These criteria are established from a fuzzy classification system and its recursive De Morgan triplet. We propose a comparison process between the classes of a fuzzy partition, based on a translation invariant similarity relation. Therefore a classification process is carried out with the equivalence relations determined by the similarity relation. Such a relation is built on the commutative group structure formed by the elements of the fuzzy classification system. Our approach is illustrated through an example on image analysis by the fuzzy c-means algorithm.

Confidence-Based Voting for the Design of Interpretable Ensembles with Fuzzy Systems

In this study, a new voting procedure for combining the fuzzy logic based classifiers and other classifiers called confidence-based voting is proposed. This method combines two classifiers, namely the fuzzy classification system, and for the cases when the fuzzy system returns high confidence levels, i.e., the returned membership value is large, the fuzzy system is used to perform classification, otherwise, the second classifier is applied. As a result, most of the sample is classified by the explainable and interpretable fuzzy system, and the second, more accurate, but less interpretable classifier is applied only for the most difficult cases. To show the efficiency of the proposed approach, a set of experiments is performed on test datasets, as well as two problems of estimating the person’s emotional state with the data collected by non-contact vital sensors, which use the Doppler effect. To validate the accuracies of the proposed approach, the statistical tests were used for comparison. The obtained results demonstrate the efficiency of the proposed technique, as it allows for both improving the classification accuracy and explaining the decision making process.

An Improved Fuzzy Classification System for Financial Credit Decision using Multi-Objective Evolutionary Optimization

In fuzzy classification system, accuracy has been gained at the cost of interpretability and vice versa. This situation is known as Interpretability-Accuracy Trade-off. To handle this trade-off between accuracy and interpretability the evolutionary algorithms (EAs) are often used to optimize the performance of the fuzzy classification system. From the last two decades, several multi-objective evolutionary systems have been designed and successfully implemented in several fields for finding multiple solutions at a single run. In Financial Decision making concerning Credit Allocation, Classification is a significant component to obtain credit scores and predict bankruptcy. A fuzzy classification system for the financial credit decision has been designed and find out the Accuracy and Interpretability parameters for applying various MOEAs to get the pareto optimal solution resulting in to improvement in the performance of the proposed system. The proposed model implemented on standard benchmark financial credit allocation datasets i.e., German Credit Approval system available from the UCI repository of machine learning databases (http://archive.ics.uci.edu/ml) and using the open source tool MOEA framework (http://www.moeaframework.org). The experimental analysis highlights that the NSGA-III works efficiently for financial credit approval system and improves the performance by making a balanced trade-off between accuracy and interpretability.