ElAr :: Перегляд за автором "Bodyanskiy, Ye. V."

Перегляд за автором "Bodyanskiy, Ye. V."

Зараз показано 1 - 13 з 13

Adaptive human machine interaction approach for feature selection-extraction task in medical data mining
(International Journal of Computing, 2018) Perova, I.; Bodyanskiy, Ye. V.
Feature Selection task is one of the most complicated and actual in the areas of Data Mining and Human Machine Interaction. Many approaches to its solving are based on non-mathematical and presentative hypothesis. New approach to evaluation of medical features information quantity, based on optimized combination of feature selection and feature extraction methods is proposed. This approach allows us to produce optimal reduced number of features with linguistic interpreting of each of them. Hybrid system of feature selection/extraction based on Neural Network-Physician interaction is investigated. This system is numerically simple, can produce feature selection/extraction with any number of factors in online mode using neural network-physician interaction based on Oja’s neurons for online principal component analysis and calculating distance between first principal component and all input features. A series of experiments confirms efficiency of proposed approaches in Medical Data Mining area and allows physicians to have the most informative features without losing their linguistic interpreting.
Adaptive Neuro-Fuzzy Methods for Distorted Data Clustering
(2020) Shafronenko, A. Yu.; Bodyanskiy, Ye. V.; Rudenko, D. A.
The problem of data sets described by vector-images clustering often occurs in many applications associated with Data Mining [40, 41], when processed vector-image with different levels of probabilities, possibilities or memberships, can belong to more than one class. However, there are situations when the data sets contain missing values. In this situation more effective is to use mathematical apparatus of Computational Intelligence [Rutkowski, 2008] and, first of all artificial neural networks [8], that solve task of restoring the lost observations and modifications of the popular method of fuzzy c-means [22], which solve the problem of clustering without recovery of data. Existing approaches for data processing with missing values [6], are efficient in cases when the massive of the original observations is given in batch form and does not change during the processing. At the same time, there is a wide class of problems in which the data that arrive to the processing, have the form of sequence that is feed in real time as it occurs in the training of Kohonen self-organizing maps [1] or their modifications [2]. In this regard we have introduced [42] the adaptive neuro-fuzzy Kohonen network to solve the problem of clustering data with gaps based on the strategy of partial distances (PDS FCM). However, in situations where the number of such missing values is too big, the strategy of partial distances may be not effective, and therefore it may be necessary, along with the solution of fuzzy clustering simultaneously estimate the missing observations. In this situation, a more efficient is approach that is based on the optimal expansion strategy (OCS FCM) [22]. This chapter is devoted to the task of on-line data clustering using the optimal expansion strategy, adapted to the case when information is processed in a sequential mode, and its volume is not determined in advance.
Evolutionary Algorithms Based on Cat Swarm Optomozation in Fuzzy Clustering Tasks
(2020) Shafronenko, A. Yu.; Bodyanskiy, Ye. V.; Rudenko, D. A.
Computational intelligence methods are widely used to solve many complex problems, both purely scientific, and in the field of engineering, business, finance, medical and technical diagnostics, and other areas related to information processing, including, of course, traditional intellectual analysis: Data Mining and such new directions as Dynamic Data Mining, Data Stream Mining, Big Data Mining, Web Mining, Text Mining, etc. [1-6]. One of the main areas of computational intelligence is evolutionary algorithms, which are essentially certain mathematical models of reproduction or development of biological organisms, inspired by nature and intended, in the most general case, to find the global optimum of multi-extremal functions under uncertainty. Historically, the first evolutionary algorithms were the so-called genetic algorithms, which are based on the selection and genetics mechanisms that implement the survival of the strongest individuals in the process of evolution. The most popular evolutionary bioinspired algorithms for today are the so-called “swarm” procedures (Particle Swarm Optimization - PSO) [7], among which, as for the performance and simplicity of implementation is Cat Swarm Optimization (CSO) [8,9]. These algorithms have confirmed their effectiveness in solving a number of rather complex tasks and have already “managed” to undergo a number of modifications, among which are procedures based on harmonic search, fractional derivatives, adaptation of search parameters and, finally, “crazy cats” [ 10-16]. At the same time, these procedures are not deprived of some shortcomings that degrade the properties of the global extremum search process. The aim of the chapter is to develop a fast-acting and numerically simple method of evolutionary optimization under conditions of multi-extremal goal functions.
Fast medical diagnostics using autoassociative neuro-fuzzy memory
(International Journal of Computing, 2017) Perova, I.; Bodyanskiy, Ye. V.
This paper proposes an architecture of fast medical diagnostics system based on autoassociative neuro-fuzzy memory. The architecture of proposed system is close to traditional Takagi-Sugeno-Kang neuro-fuzzy system, but it is based on other principles. This system contains of recording subsystem and pattern retrieval subsystem, where diagnostics of patients with unknown diagnoses is realized. Level of memberships for all other possible diagnoses from recording subsystem is determined too. System tuning is based on lazy learning procedure and “neurons in data points” principle and uses bell-shaped fuzzy basis functions. Number of these functions changes during training process using principles of evolving connectionist systems. Bell-shaped membership functions centers can be tuned using proposed algorithm, processes of accumulation patients in fundamental memory and patients retrieval are described. This hybrid neuro-fuzzy associative memory combines advantages of fuzzy inference systems, artificial neural networks and evolving systems and its using provides the increasing of autoassociative memories capacity without essential complication of its architecture for medical diagnostics tasks.
Kernel principal component analysis in data stream mining tasks
(2016) Bodyanskiy, Ye. V.; Deineko, A. O.; Eze, F. M.; Shalamov, M. O.
Currently, self-learning systems of computational intelligence [1, 2] and, above all , artificial neural networks (ANN ), that tune their parameters without a teacher on the basis of the self-learning paradigm [3], are widely used in solving various problems of Data Mining, Exploratory Data Analysis etc. Among these tasks, most frequently encountered in the Text Mining, Web Mining, Medical Data Mining, it be can mentioned the problem of compression of large data sets, for whose solution principal component analysis (PCA) is widely used, which consists in the orthogonal projection of input data vectors from the original n-dimensional space in the m- dimensional space of reduced dimensionality
Medical online neuro-fuzzy diagnostics system with active learning
(International Journal of Advances in Computer and Electronics Engineering, 2017) Bodyanskiy, Ye. V.; Perova, I.
Situations when in the medical data set some patients have known diagnoses and all other have unknown ones is spread wise problem of present-day medicine. Known systems of computational intelligence show mediocre level of diagnostics in these data sets. In this paper online neuro-fuzzy diagnostics system with active learning is proposed. This system allows to increase a quality of medical diagnostics under the condition of small number of known reference signals due to combination of special learning algorithms – active learning. The proposed online neuro-fuzzy system is based on popular neural networks as Self-Organizing Map (SOM) and Learning Vector Quantization network (LVQ). Active learning procedure permits to tune their synaptic weights using simple recurrent self-learning procedures (SOM) and controlled learning with teacher (LVQ). Neuro-fuzzy diagnostics system with active learningwas used for breast cancer in Wisconsin data set processing and showed higher level of classification-clusterization results comparatively with known systems
On-line robust fuzzy clustering based on similarity measure
(2013) Bodyanskiy, Ye. V.; Shafronenko, A. Yu.
The problem offuzzy adaptive on-line clustering of data distorted by outliers sequentially supplied to the processing when the original sample volume and the number of distorted observations are unknown is considered. The probabilistic and possibilistic clustering algorithms for such data, that are based on the similarity measure of a special kind that weakens or overwhelming outliers are proposed.
Online algorithm for possibilitic fuzzy clustering based on evolutionary cat swarm optimization
(2019) Bodyanskiy, Ye. V.; Shafronenko, A. Yu.
The problem of clustering of multidimensional observations is often found in many applications related to data mining and exploratory data analysis. The traditional approach to solving these problems requires that every observation could belong to only one cluster at a more natural is situations when a feature vector with the various possible levels of memberships can belong to multiple classes. This situation is the subject of fuzzy cluster analysis, rapidly developing now. We propose online adaptive approach for this task solving.
Online fuzzy clustering of high dimension data streams based on neural network ensembles
(Innovative Technologies and Scientific Solutions for Industries, 2019) Bodyanskiy, Ye. V.; Perova, I.; Zhernova, P.
The subject matter of the article is fuzzy clustering of high-dimensional data based on the ensemble approach, provided that a number and shape of clusters are not known. The goal of the work is to create the neuro-fuzzy approach for clustering data when the data stream is fed for online processing and a number and shape of clusters are unknown. The following tasks are solved in the article - the input feature space is compressed in the online mode; the model of neural network ensembles for data clustering is built; the ensemble of neuro-fuzzy networks for clustering high-dimensional data is developed; the approach for clustering data in the online mode is worked out. The following results are obtained - the main idea of the proposed approach is based on a modification of the fuzzy C-means algorithm. To reduce the dimension of the input space, the modified Hebb-Sanger network is suggested to be used; this net is characterized by the increased speed and is built on the basis of the modified Oja neurons. A speed-optimized learning algorithm for the Oja neuron is proposed. Such a network implements the method of principal components in the online mode with high speed. Conclusions. In the event the reduction-compression procedure cannot be used due to the probability of losing the physical meaning of the original space, a new clustering criterion was introduced; this criterion contains both a well-known polynomial fuzzifier and the weighment of individual components of the deviations of presented images from cluster centroids. The recurrent modification based on the algorithms proposed in this article is introduced. A mathematical model is developed to determine the quality of clustering with the use of the Xi-Beni index, which was modified for the online mode. The experimental results confirm the fact that the proposed system enables solving a wide range of Data Mining tasks when data sets are processed online, provided that a number and shape of clusters are not known and there is a large number of observations as well.
Online Medical Data Stream Mining Based on Adaptive Neuro-Fuzzy Approaches
(Cybernetics and computer engineering, 2019) Perova, I.; Bodyanskiy, Ye. V.
Data mining approaches in medical diagnostics tasks have a number of special properties that do not allow the use of such approaches in a classical form. That’s why adaptive neuro-fuzzy systems for online medical data stream processing tasks and its learning algorithms have been developed. Proposed systems can process medical data streams in three modes: supervised learning, unsupervised learning and active learning. The purpose of the paper is to develop approach, based on adaptive neuro-fuzzy systems to solve the tasks of medical data stream mining in online-mode.
Online Recurrent Me-thod Of Credibilistic Fuzzy Clusterin
(2020) Bodyanskiy, Ye. V.; Shafronenko, A. Yu.; Rudenko, D. A.; Klymova, I. N.
An online method of reliable fuzzy clustering is proposed, designed to analyze data sequentially received for processing. A feature of the developed approach is the use of the membership function of a special kind described by the density function of the Cauchy distribution. The actual procedure for clarifying the centroids of clusters is essentially a self-learning rule “The Winner Takes More” (WTM), in which the neighborhood function is generated by the introduced membership function.
Online Сredibilistic Fuzzy Clustering of Data with Gaps
(2020) Bodyanskiy, Ye. V.; Shafronenko, A. Yu.; Klymova, I. N.
The problem of online fuzzy clustering based on credibilistic approaches based on the partial distance strategy and processing of data with gaps was proposed. The considered methods are simple in numerical implementation, being essentially gradient procedures for optimizing objective functions of a special type.
The Fast Modification of Evolutionary Bioinspired Cat Swarm Optimization Method
(2019) Shafronenko, A.Yu.; Bodyanskiy, Ye. V.; Pliss, I. P.
This paper discusses the optimization problem based on cat swarm optimization by introducing elements of a random search in a stochastic modification of the basic procedure into the seeking and tracing modes, that improves the speed and accuracy of determining the direction of movement in the seeking mode and improves the global properties of the procedure in the tracing mode. The proposed optimization method, being a representative of evolutionary algorithms, is intended for using hybrid systems of computational intelligence and, above all, in learning tasks of artificial neural networks, neuro-fuzzy systems, as well as in clustering and classification problems.

Перегляд за автором "Bodyanskiy, Ye. V."

Результатів на сторінку

Варіанти сортування