Neuro-fuzzy Classifier Research Articles

Problems in pattern classification in high dimensional spaces: behavior of a class of combined neuro-fuzzy classifiers

The designer of a pattern classification system is often faced with the following situation: finite sets of samples, from the various classes, are available along with a large set of measurements, or features, to be computed from the patterns. In pattern recognition literature, several investigations have demonstrated the relationship between dimensionality, sample size and recognition accuracy. The ultimate goal of this paper is to study the generalization error of statistical (1_NN, 4_NN), fuzzy (FCM, FKCN), neural (MLPNN, RBFNN) and neuro-fuzzy classifiers in high dimensional spaces. Computational complexity of classification algorithms in high dimensional spaces is also discussed. Our experimental results show the robustness of fuzzy, neural, and neuro-fuzzy classifiers to the curse of dimensionality.

Adaptive resolution min-max classifiers

A high automation degree is one of the most important features of data driven modeling tools and it should be taken into consideration in classification systems design. In this regard, constructive training algorithms are essential to improve the automation degree of a modeling system. Among neuro-fuzzy classifiers, Simpson's (1992) min-max networks have the advantage of being trained in a constructive way. The use of the hyperbox, as a frame on which different membership functions can be tailored, makes the min-max model a flexible tool. However, the original training algorithm evidences some serious drawbacks, together with a low automation degree. In order to overcome these inconveniences, in this paper two new learning algorithms for fuzzy min-max neural classifiers are proposed: the adaptive resolution classifier (ARC) and its pruning version (PARC). ARC/PARC generates a regularized min-max network by a succession of hyperbox cuts. The generalization capability of ARC/PARC technique mostly depends on the adopted cutting strategy. By using a recursive cutting procedure (R-ARC and R-PARC) it is possible to obtain better results. ARC, PARC, R-ARC, and R-PARC are characterized by a high automation degree and allow to achieve networks with a remarkable generalization capability. Their performances are evaluated through a set of toy problems and real data benchmarks. The paper also proposes a suitable index that can be used for the sensitivity analysis of the classification systems under consideration.

FAULT DIAGNOSIS OF AN INDUSTRIAL GAS TURBINE USING NEURO-FUZZY METHODS

The paper focuses on the application of neuro-fuzzy techniques in fault detection and isolation. The objective of this paper is to detect and isolate faults to an industrial gas turbine, with emphasis on faults occurred in the actuator part of the gas turbine. A neuro-fuzzy based learning and adaptation of TSK fuzzy models is used for residual generation, while for residual evaluation a neuro-fuzzy classifier for Mamdani models is used. The paper is concerned on how to obtain an interpretable fault classifier as well as interpretable models for residual generation.

Proposed wavelet-neurofuzzy combined system for power quality violations detection and diagnosis

A system for the identification of power quality violations is proposed. It is a two-stage system that employs the potentials of the wavelet transform and the adaptive neurofuzzy networks. For the first stage, the wavelet multiresolution signal analysis is exploited to denoise and then decompose the monitored signals of the power quality events to extract its detailed information. A new optimal feature-vector is suggested and adopted in learning the neurofuzzy classifier. Thus, the amount of needed training data is extensively reduced. A modified organisation map of the neurofuzzy classifier has significantly improved the diagnosis efficiency. Simulation results confirm the aptness and the capability of the proposed system in power quality violations detection and automatic diagnosis.

Pattern classification by a neurofuzzy network: application to vibration monitoring

An innovative neurofuzzy network is proposed herein for pattern classification applications, specifically for vibration monitoring. A fuzzy set interpretation is incorporated into the network design to handle imprecise information. A neural network architecture is used to automatically deduce fuzzy if-then rules based on a hybrid supervised learning scheme. The neurofuzzy classifier proposed is equipped with a one-pass, on-line, and incremental learning algorithm. This network can be considered a self-organized classifier with the ability to adaptively learn new information without forgetting old knowledge. The classification performance of the proposed neurofuzzy network is validated on the Fisher's Iris data, which is a well-known benchmark data set. For the generalization capability, the neurofuzzy network can achieve 97.33% correct classification. In addition, to demonstrate the efficiency and effectiveness of the proposed neurofuzzy paradigm, numerical simulations have been performed using the Westland data set. The Westland data set consists of vibration data collected from a US Navy CH-46E helicopter test stand. Using a simple fast Fourier transform technique for feature extraction, the proposed neurofuzzy network has shown promising results. Using various torque levels for training and testing, the network achieved 100% correct classification.

Classification of Hydrometeors Based on Polarimetric Radar Measurements: Development of Fuzzy Logic and Neuro-Fuzzy Systems, and In Situ Verification

A fuzzy logic and neuro-fuzzy system for classification of hydrometeor type based on polarimetric radar measurements is described in this paper. The hydrometeor classification system is implemented by using fuzzy logic and a neural network, where the fuzzy logic is used to infer hydrometeor type, and the neural network learning algorithm is used for automatic adjustment of the parameters of the fuzzy sets in the fuzzy logic system according to prior knowledge. Five radar measurements, namely, horizontal reflectivity ( ZH), differential reflectivity (ZDR), differential propagation phase shift ( KDP), correlation coefficient [rHV(0)], and linear depolarization ratio (LDR), and corresponding altitude, have been used as input variables to the neuro-fuzzy network. The output of the neuro-fuzzy system is one of the many possible hydrometeor types: 1) drizzle, 2) rain, 3) dry and low density snow, 4) dry and high-density crystals, 5) wet and melting snow, 6) dry graupel, 7) wet graupel, 8) small hail, 9) large hail, and 10) a mixture of rain and hail. The neuro-fuzzy classifier is more advantageous than a simple neural network or a fuzzy logic classifier because it is more transparent (instead of a ‘‘black box’’) and can learn the parameter of the system from the past data (unlike a fuzzy logic system). The hydrometeor classifier has been applied to several case studies and the results are compared against in situ observations.

A New Neuro-Fuzzy Classifier with Application to On-Line Face Detection and Recognition

In this paper, we propose a neuro-fuzzy classifier (NEFCAR) that utilizes positive and negative rules with different rule importances to create the decision boundaries between different classes. The locally unsupervised and globally supervised training technique is adopted. The decision-based and approximation-based strategies are combined to provide a suitable amount of training for each training pattern. The reinforced and anti-reinforced learning rules are given with different weighting so that the training can be efficient and can reach convergence quickly. Moreover, NEFCAR can easily provide the confidence measure of each classification decision. Therefore, the rejection algorithm can be implemented in a straightforward manner. Noise tolerant training is conducted to improve the generalization performance and the confidence measure is adopted to avoid overtraining. The proposed classifier is applied to two applications. The first one is the Fisher iris data classification, and the second one is an on-line face detection and recognition application. Good classification results are obtained in both applications. In the on-line face detection and recognition system, two NEFCAR's are utilized: a two-class and a multi-class NEFCAR's are adopted to detect the face and recognize the face, respectively. The color of skin and the motion information are taken into consideration heuristically to improve the effectiveness of the face location algorithm.

Hybrid modeling approach to on-line estimation of yeast biomass concentration in industrial bioreactor

The industrial fed-batch yeast cultivation process has been divided into four different metabolic phases (adaptation, carbon limited, oxygen limited and maturation) by a neuro-fuzzy classification model that consists of 4 applied linguistic rules on 2 state variables: oxygen uptake rate and liquid volume. The membership functions have been automatically adapted by this fuzzy perceptron, i.e., by a supervised learning algorithm initialized by prior operator's knowledge. Process compartmentalization has made easier and more realistic a subsequent state estimation of the biomass concentration with separate artificial neural networks combined with balance equations. Static networks with local recurrent memory structures were used, and the inputs were standard cultivation state variables: respiratory quotient, molasses feed rate, ethanol concentration, etc. This hybrid approach is generally applicable to state estimation or prediction when different sources of process information and knowledge have to be integrated.

Neuro-fuzzy approach versus rough-set inspired methodology for intelligent decision support

One of the two goals of this paper is to briefly present two different methodologies that can be used to the design of intelligent decision support systems, in particular, from the field of medicine. The first approach, combining artificial neural networks and fuzzy sets, yields a neuro-fuzzy classifier that can be trained with both purely numerical data as well as qualitative, linguistic, fuzzy data that describe the decision-making process. The second approach – resulting in a rough classifier – combines all positive aspects of rule induction systems with the flexibility of statistical techniques for classification. The second goal of this paper is to perform a broad comparative analysis of both proposed methodologies (and two others) applied to: (a) the problem of selecting surgical and non-surgical cases in the veterinary domain of equine colic, (b) the problem of diagnosing benign and malign types of breast cancer, and (c) the problem of corporate bankruptcy prediction (corporate `financial health'). Several aspects of comparison have been considered including the accuracy of the systems, diversity of the data processed, transparency and the form of decisions made.

Decision model formulation of subjective classification problem-solving knowledge using a neuro-fuzzy classifier and its effectiveness

Decision support systems developers face difficulties in formulating decision models that represent decision makers’ subjective problem-solving knowledge. Decision makers expect that formulated models are easy to understand and that those models incorporate their subjectivity in decision making. Although it is understood that knowledge discovery techniques can be used to support the processes of model formulation, there is little empirical evidence on the effectiveness of such approaches. In this paper, we investigate the effectiveness of a knowledge discovery technique in automated modeling of subjective classification problem-solving knowledge. We identify two measures of agreement and similarity to evaluate the effectiveness of a neuro-fuzzy classifier in this task. Findings from the experiment show that the neuro-fuzzy classifier can effectively model subjective classification problem-solving knowledge from a small set of examples, and it can represent the formulated models as a concise and intuitive set of fuzzy rules.

Obtaining interpretable fuzzy classification rules from medical data

For many application problems classifiers can be used to support a decision making process. In some domains-in areas like medicine especially-it is preferable not to use black box approaches. The user should be able to understand the classifier and to evaluate its results. Fuzzy rule based classifiers are especially suitable, because they consist of simple linguistically interpretable rules and do not have some of the drawbacks of symbolic or crisp rule based classifiers. Classifiers must often be created from data by a learning process, because there is not enough expert knowledge to determine their parameters completely. A simple and convenient way to learn fuzzy classifiers from data is provided by neuro-fuzzy approaches. In this paper we discuss extensions to the learning algorithms of neuro-fuzzy classification (NEFCLASS), a neuro-fuzzy approach for data analysis that we have presented before. We present interactive strategies for pruning rules and variables from a trained classifier to enhance its readability, and demonstrate our approach on a small example.

NEFCLASS-X — a Soft Computing Tool to Build Readable Fuzzy Classifiers

Neuro-fuzzy classification systems offer a means of obtaining fuzzy classification rules by a learning algorithm. Although it is usually no problem to find a suitable fuzzy classifier by learning from data, it can, however, be hard to obtain a classifier that can be interpreted conveniently. There is usually a trade-off between accuracy and readability. This paper discusses NEFCLASS — a neuro-fuzzy approach for classification problems — and its implementation NEFCLASS-X. It is shown how a readable fuzzy classifier can be obtained by a learning process and how interactive strategies for pruning rules and variables from a trained classifier can enhance its interpretability.

Supervision of vehicles' tyre pressures by measurement of body accelerations

Due to the rising consciousness of safety aspects, the supervision of vehicle's tyre pressures is a major aspect of improved active car safety. Therefore, in this paper a method for monitoring the tyre pressures is presented, using body acceleration signals. Analyzing the frequency spectrum of the virtual transfer function between the body acceleration at the front and rear wheel on one side of the vehicle, characteristic features are generated. Thereby, external interferences on the spectrum and their influences on the symptoms are discussed. To quantify the tyre pressure, a neuro-fuzzy classification of the characteristics is applied.

Estimating the crowding level with a neuro-fuzzy classifier

This paper introduces a neuro-fuzzy system for the estimation of the crowding level in a scene. Monitoring the number of people present in a given indoor environment is a requirement in a variety of surveillance applications. In the present work, crowding has to be estimated from the image processing of visual scenes collected via a TV camera. A suitable preprocessing of the images, along with an ad hoc feature extraction process, is discussed. Estimation of the crowding level in the feature space is described in terms of a fuzzy decision rule, which relies on the membership of input patterns to a set of partially overlapping crowding classes, comprehensive of doubt classifications and outliers. A society of neural networks, either multilayer perceptrons or hyper radial basis functions, is trained to model individual class-membership functions. Integration of the neural nets within the fuzzy decision rule results in an overall neuro-fuzzy classifier. Important topics concerning the generalization ability, the robustness, the adaptivity and the performance evaluation of the system are explored. Experiments with real-world data were accomplished, comparing the present approach with statistical pattern recognition techniques, namely linear discriminant analysis and nearest neighbor. Experimental results validate the neuro-fuzzy approach to a large extent. The system is currently working successfully as a part of a monitoring system in the Dinegro underground station in Genoa, Italy.

Supervision of Vehicles' Tire Pressure by Measurement of Body Accelerations

Due to the rising consciousness of safety aspects the supervision of vehicles' tire pressure is a major effort to improve active car safety. Therefore, in this contribution a method for monitoring the tire pressure is presented using body acceleration signals. Analysing the frequency spectrum of the virtual transfer function between the body acceleration at the front and the rear wheel of one side of the vehicle characteristic features are generated. Thereby, external interferences to the spectrum and their influences to the symptoms are discussed. Then, a neuro-fuzzy classification of the characteristics is applied to quantify the tire pressure.

A neuro-fuzzy system for chemical agent detection

The authors previously introduced a fuzzy version of Kohonen's well-known self-organizing map neural network model. In this novel neuro-fuzzy system, the neurons of Kohonen's original model are replaced by fuzzy rules. Each fuzzy rule is composed of fuzzy sets and an output singleton. Since the fuzzy self-organizing map is a modified version of Kohonen's original model, the self-organizing map and the learning vector quantization learning laws can be used to tune the neuro-fuzzy system. Originally, the fuzzy self-organizing map was intended to be used as an unknown function approximator, while Kohonen's self-organizing map is primarily used as a neural classifier. In this paper, the authors show how the fuzzy self-organizing map can also be used as a neuro-fuzzy classifier. Simulation results show that, in chemical agent detection, the fuzzy self-organizing map not only gives better classification results than Kohonen's model, but it also has smaller number of fuzzy rules than the corresponding neurons required by Kohonen's self-organizing map.

ON BUILDING A KOHONEN NEURAL NET PARALLEL SIMULATOR∗

This paper presents a Kohonen neural net parallel simulator. The simulator was developed on a Sequent Balance 8000 computer system. Comparative results emphasize the impact of the different strategies of parallelization, the number of processors involved and inter-processor communications over the efficiency of the parallel implementation. The simulator was used in a pattern recognition application, in the automatic synthesis of a neuro-fuzzy classifier. The results obtained in the neuro-fuzzy classifier synthesis and in the classification process are presented. Some further work proposals, the discussions and the conclusions complete the paper.