Estimation of the Probability in the System of "Classifiers

Le Centre pour la Communication Scientifique Directe - HAL - Archive ouverte HAL, 2015

HAL is a multidisciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

international conference on adaptive and natural computing algorithms, 1998

The main aim of this paper is three fold: a) to understand the working of a classifier system based on possibility distribution functions, b) to evaluate its performance against other superior methods such as fuzzy and non-fuzzy neural networks on real data, c) and finally to recommend changes for enhancing its performance. The paper explains how to construct a possibility based classifier system which is used with conventional error-estimation techniques such as crossvalidation and boot-strapping. The results were obtained on a set of electronic nose data and this performance was compared with earlier published results on the same data using fuzzy and non-fuzzy neural networks. The results show that the possibility approach is superior to the non-fuzzy approach, however, further work needs to be done.

Studies in Fuzziness and Soft Computing, 2000

6. Training of fuzzy if-then classifiers 6.1 Expert opinion or data analysis? Assume we have a difficult pattern recognition problem which can easily be handled by a human but not by a machine. Assume also that the human recognition process is difficult to articulate or express in any functional or algorithmic form. Examples of such tasks are face recognition and speaker verification. In some problems we have some knowledge about the classes. An example is handwriting recognition where the theoretical shapes, connections, loops, etc. for each symbol are known, so the "ideal" prototype for each class is described by a set of rules. Nevertheless, handwriting recognition by a machine (and sometimes by a human) is stiH a challenge. Two natural approaches to designing a classifier are • Ask an expert how they solve the problem and try to encapsulate the knowledge in a fuzzy rule-base classifier. • Collect input-output data (Le., a labeled data set) and extract the classifier parameters from the data. While the first model is interpretable in the domain context (the classifier is said to be transparent), the model based on data may or may not be interpretable. Fuzzy models are deemed to be able to integrate both human and data sources. There is no unique way for this integration, so the literature offers a whole spectrum of possible fuzzy classifier designs ranging from entirely expert-driven to entirely data-driven ones. Data-driven classifiers are usually more accurate and less interpretable than expert-driven classifiers. Practice has given rise to the interpretability-accuracy dilemma [33, 301). It seems that the two characteristics have a compensatory behavior. Highly accurate classifiers usually need a sophisticated classification mechanism (e.g., many if-then rules, intricate formulas, etc.) which cannot easily be translated into the domain jargon. Most of the recent effort in fuzzy classifier design is focused on deriving the classifier from data. There are studies that look at a combination of expert and data sources in classifier design (e.g., [144, 338)) but we will confine this chapter to data-driven fuzzy designs only.

1992

Difficult pattern recognition problems involving large class sets and noisy input can be solved by a multiple classifier system, which allows simultaneous use of arbitrary feature descriptors and classification procedures. Independent decisions by each classifier can be combined by methods of the highest rank, Borda count, and logistic regression, resulting in substantial improvement in overall correctness

We present a scheme for object recognition by classificatory problem solving in the framework of fuzzy sets and possibility theory. The scheme has a particular focus on handling of the imperfection problems that are common in application domains where the objects to be recognized (detected and identified) represent undesirable situations, referred to as crises. Crises develop over time and observations typically increases in number and precision as the crisis develops. Early detection and precise recognition of crises is desired, since it increases the possibility of an effective treatment. The crisis recognition problem is central in several areas of decision support, such as in medical diagnosis, financial decisionmaking, and early warning systems. The problem is characterized by vague knowledge and observations suffering from several kinds of imperfections, such as missing information, imprecision, uncertainty, unreliability of the source, and mutual, possible conflicting or reinforcing observations of the same phenomena. The problem of handling possibly imperfect observations from multiple sources includes the problems of information fusion and multiple sensor data fusion. The different kinds of imperfection are handled in the framework of fuzzy sets and possibility theory.

2002

Bayes Classifiers are widely used currently for recognition, identification and knowledge discovery. The fields of application are, for example, image processing, medicine, chemistry (QSAR). But by mysterious way the Naive Bayes Classifier usually gives a very nice and good presentation of a recognition. It can not be improved considerably by more complex models of Bayes Classifier. We demonstrate here a very nice and simple proof of the Naive Bayes Classifier optimality, that can explain this interesting fact.The derivation in the current paper is based on arXiv:cs/0202020v1

As multiple experts can confront and exchange their ideas in order to improve the decision-making process, a pattern recognition system can use several classifiers in order to improve its recognition rate. Moreover, various decisions strategies, implying these classifiers in different ways, can contribute to a same recognition task. A first strategy consists in deciding using different opinions: this is the combination of classifiers. A second strategy consists in using one or more opinions for better guiding other classifiers in their training stages, and/or to improve the decision-making of other classifiers in the classification stage: this is the cooperation of classifiers. The third and last strategy consists in giving more importance to one or more classifiers according to various criteria or situations: this implies the selection of classifiers. Since time could be a selection criterion, the temporal aspect of pattern recognition, i.e. the possible evolution of the classes to...

Principles of Data Mining

1997

In the last years, great attention has been devoted to multiple classifier systems. The implementation of such a system implies the definition of a rule (combining rule) for determining the most likely class, on the basis of the class attributed by each single expert. The availability of a criterion to evaluate the credibility of the decision taken by a classifier can be profitable in order to implement the combining rule. We propose a method that, after defining the reliability of a classification on the basis of information directly derived from the output of the classifier, uses this information in the context of a combining rule. The results obtained by combining four handwritten character recognizers on the basis of classification reliability are compared with those obtained by using three different combining criteria. Tests have been performed using a standard handwritten character database.

2011

In this paper we consider majority voting of multiple classifiers systems in the case of two-valued decision support for many-class problem. Using an explicit representation of the classification error probability for ensemble binomial voting and two class problem, we obtain general equation for classification error probability for the case under consideration. Thus we are extending theoretical analysis of the given subject initially performed for the two class problem by Hassen and Salamon and still used by Kuncheva and other researchers. This allows us to observe important dependence of maximal posterior error probability of base classifier allowable for building multiple classifiers from the number of considered classes. This indicates the possibility of improving the performance of multiple classifiers for multiclass problems, which may have important implications for their future applications in many fields of science and industry, including the problems of machines diagnostic and systems reliability testing.