The data-driven identification of fuzzy rule-based classifiers for high-dimensional problems is addressed. A binary decision-tree-based initialization of fuzzy classifiers is proposed for the selection of the relevant features and e#ective initial partitioning of the input domains of the fuzzy system. Fuzzy classifiers have more flexible decision boundaries than decision trees (DTs) and can therefore be more parsimonious. Hence, the decision tree initialized fuzzy classifier is reduced in an iterative scheme by means of similarity-driven rule-reduction. To improve classification performance of the reduced fuzzy system, a genetic algorithm with a multi-objective criterion searching for both redundancy and accuracy is applied. The proposed approach is studied for (i) an artificial problem, (ii) the Wisconsin Breast Cancer classification problem, and (iii) a summary of results is given for a set of well-known classification problems available from the Internet: Iris, Ionospehere, Glass, Pima, and Wine data.

Data mining consists of extracting interesting knowledge from data. This paper addresses the discovery of knowledge in the form of prediction IF-THEN rules, which are a popular form of knowledge representation in data mining. In this context, we propose a Genetic Algorithm (GA) designed specifically to discover interesting fuzzy prediction rules. The GA searches for prediction rules that are interesting in the sense of being new and surprising for the user. This is done adapting a technique little exploited in the literature, which is based on userdefined general impressions (subjective knowledge). More precisely, a prediction rule is considered interesting (or surprising) to the extent that it represents knowledge that not only was previously unknown by the user but also contradicts his original believes. In addition, the use of fuzzy logic helps to improve the comprehensibility of the rules discovered by the GA. This is due to the use of linguistic terms that are natural for the user. A prototype was implemented and applied to a real-world science & technology database, containing data about the scientific production of researchers. The GA implemented in this prototype was evaluated by comparing it with the J4.8 algorithm, a variant of the well-known C4.5 algorithm. Experiments were carried out to evaluate both the predictive accuracy and the degree of interestingness (or surprisingness) of the rules discovered by both algorithms. The predictive accuracy obtained by the proposed GA was similar to the one obtained by J4.8, but

The automatic design of fuzzy rule-based classification systems based on labeled data is considered. It is recognized that both classification performance and interpretability are of major importance and effort is made to keep the resulting rule bases small and comprehensible. For this purpose, an iterative approach for developing fuzzy classifiers is proposed. The initial model is drived from the data and subsequently, feature selection and rule base simplification are applied to reduce the model, while a Genetic Algorithm is used for parameter optimization. An application to the Wine data classification problem is shown.

In this paper, we derive new neuro-fuzzy structures called flexible neuro-fuzzy inference systems or FLEXNFIS. Based on the input–output data, we learn not only the parameters of the membership functions but also the type of the systems (Mamdani or logical). Moreover, we introduce: 1) softness to fuzzy implication operators, to aggregation of rules and to connectives of antecedents; 2) certainty weights to aggregation of rules and to connectives of antecedents; and 3) parameterized families of T-norms and S-norms to fuzzy implication operators, to aggregation of rules and to connectives of antecedents. Our approach introduces more flexibility to the structure and design of neuro-fuzzy systems. Through computer simulations, we show that Mamdani-type systems are more suitable to approximation problems, whereas logical-type systems may be preferred for classification problems.

This paper presents an effective and efficient approach for translating fuzzy classification rules that use approximative sets to rules that use descriptive sets and linguistic hedges of predefined meaning. It works by first generating rules that use approximative sets from training data and then translating the resulting approximative rules into descriptive ones. Hedges that are useful for supporting such translations are provided. The translated rules are functionally equivalent to the original approxi- matire ones, or a close equivalent given search time restrictions, while reflecting their underlying preconceived meaning. Thus, fuzzy descriptive classifiers can be obtained by taking advantage of any existing approach to approximative modeling which is generally efficient and accurate, whilst employing rules that are comprehensible to human users. Experimental results are provided and comparisons to alternative approaches given.

The need and importance of automatically recognizing emotions from human speech has grown with the increasing role of human-computer interaction applications. This paper explores the detection of domain-specific emotions using a fuzzy inference system to detect two emotion categories, negative and nonnegative emotions. The input features are a combination of segmental and suprasegmental acoustic information; feature sets are selected from a 21-dimensional feature set and applied to the fuzzy classifier. Our fuzzy inference system is designed through a data-driven approach. The design of the fuzzy inference system has two phases: one for initialization for which fuzzy cmeans method is used, and the other is fine-tuning of parameters of the fuzzy model. For fine-tuning, a well known neurofuzzy method are used. Results from on spoken dialog data from a call center application show that the optimized FIS with two rules (FIS-2) improves emotion classification by 63.0 % for male data and 73.7 % for female over previous results obtained using linear discriminant classifier. 1.

A number of soft computing approaches, such as neural networks, evolutionary algorithms, and fuzzy logic, have been widely used for classifier agents to adaptively evolve solutions on classification problems. However, most work in the literature focuses on the learning ability of individual classifier agent. This paper explores incremental, collaborative learning in a multi-agent environment. We use genetic algorithm (GA) and incremental genetic algorithm (IGA) as the main techniques to evolve the rule set for classification, and employ new class acquisition as a typical example to illustrate the incremental, collaborative learning capability of classifier agents. Benchmark data sets are used to evaluate proposed approaches. The results show that GA and IGA can be successfully used for collaborative learning among classifier agents.

The paper proposes a new Empirical Risk Functional as cost function for training neuro-fuzzy classifiers. This cost function, called Approximate Differentiable Empirical Risk Functional (ADERF), provides a differentiable approximation of the misclassification rate so that the Empirical Risk Minimization Principle formulated in Vapnik's Statistical Learning Theory can be applied. Also, based on the proposed ADERF, a learning algorithm is formulated. Experimental results on a number of benchmark classification tasks are provided and comparison to alternative approaches given.

In recent years, we have witnessed a strong emphasis on high performance and precision of fuzzy systems. Many publications are focused on data driven approaches, i.e., the construction of fuzzy systems from data and applying them in areas like data mining, pattern recognition, prediction or control. In such applications, fuzzy system inevitably must be compared with other inductive methods, like neural networks, machine learning or statistical techniques. The most prominent feature that distinguishes fuzzy systems from many other techniques is their transparency and interpretability. Fuzzy models are ideally suited for explaining solutions to users. In the current literature, however, surprisingly little attention has been devoted to the study of the interplay between interpretability and precision. These objectives are to a certain degree conflicting and attention must be paid to both of them. In this paper, the interpretation and transparency issues are first discussed with regard to the various parameters (degrees of freedom) in the Mamdani and Takagi--Sugeno models. An overview is also given of methods for improving the transparency and interpretability of fuzzy systems induced from data. These include the use of similarity measures, semantic constraints and multi-objective optimization.

The trace element content of clinkers (and possibly of cements) can be used for the qualitative identification (i.e. manufacturing factory). This paper proposes a fuzzy classifier for the discrimination of clinkers produced in different factories based on their Mg, Sr, Ba, Mn, Ti, Zr, Zn and V content. The fuzzy classifier is identified by unsupervised fuzzy clustering. The most relevant trace elements were selected based on the obtained clusters by the modified version of Fisher interclass separability method. The classification of Portuguese and South African clinkers is used as an illustrative example. The results show that the proposed method is useful to identify compact classifiers that are able to determine the origin of the clinker; and the obtained classifier is easy to use and interpret for engineers and researchers, even when they are not familiar with the concept of fuzzy logic.