Fuzzy Neural Network Model Research Articles

An advanced interpretable Fuzzy Neural Network model based on uni-nullneuron constructed from n-uninorms

This paper formulates a fuzzy logic neuron that uses n-uninorms to construct uni-nullneurons. A fuzzy neural network (FNN) composed of these neurons is easy to operate with nullnorms and uninorms at different times, which results in high accuracy of the model outputs and increases the flexibility in connecting the rule antecedents (enabling AND and OR connections within one rule). This, in turn, may allow experts/operators to extract more knowledge from data. The FNN uses a one-versus-rest classifier learning scheme for multi-class classification problems, where neuron activation levels construct the (indicator) regression matrix; this results in a non-linear regression by indicator, which can resolve the inherent class masking problem apparent in the linear case. We propose a specific neuron-selection strategy in the learning stage that applies Lasso to bootstrap replications in order to ensure that the rule base is as compact as possible and induced by a low number of neurons. To evaluate the new neuron acting in FNNs, we performed pattern classification, and regression tests. Compared with traditional FNN in the literature, our variant showed improved model accuracies for several high-dimensional real-world datasets in binary and multi-class classification and regression problems. Combined with the ability to generate human-readable rules, this offers the ability to generate parsimonious responses with a high degree of confidence.

Determining the Composition Functions of Persian Non-standard Sentences in Terminology using a Deep Learning Fuzzy Neural Network Model

Organizations can enhance the speed of well-informed decision-making by correctly understanding and using data. Since there is a tremendous gap between the speed of data processing and data generation in the world, exploring data mining in the digital world becomes inevitable. In the Persian language, similar to other languages, with the expansion of communications through social networks, the spelling of words has become abridged and the engagement of foreign loan words and emoticons has been increasing on a daily basis. Given the richness of Persian and its typographical-grammatical similarities to Arabic, research in Persian can be applied to other akin languages as well. In this regard, the current study deals with data mining of Persian non-standard sentences in order to find the function of each word in the sentence. The volume of computation might be limited in traditional methods of natural language processing for each factor contributing to functions. That is because the minimum number of computations is (5 × number of words 9) + (5 × number of words 15). Therefore, this study adopted the Gated Recurrent Unit (GRU) method to process such computations. The newly proposed method reinforces the results of word function identification by using two categories of independent and dependent Persian language functions as well as five factors contributing to the functions of words in sentences as five output gates. Meanwhile, the values of the training tables in this method are fuzzy, where the center-of-gravity method is adopted to decide on the values as well as to reduce the complexity and ambiguity of such computations on the probability of each event occurring. Therefore, the new method is briefly called fuzzy GRU. The results show that the proposed algorithm achieves 80 % reduction in the amount of calculations per gate of updates and reinforcement is approximately 2 % up from 67 % in standard sentences to 69 % of the non-standard sentences.

Bayesian Stochastic Neural Network Model for Turbomachinery Damage Prediction

Turbomachinery often suffers various defects such as impeller cracking, resulting in forced outage, increased maintenance costs, and reduced productivity. Condition monitoring and damage prognostics has been widely used as an increasingly important and powerful tool to improve the system availability, reliability, performance, and maintainability, but still very challenging due to multiple sources of data uncertainties and the complexity of analytics modeling. This paper presents an intelligent probabilistic methodology for anomaly prediction of high-fidelity turbomachine, considering multiple data imperfections and multivariate correlation. The proposed method adeptly integrates several advanced state-of-the-art signal processing and artificial intelligence techniques: wavelet multi-resolution decomposition, Bayesian hypothesis testing, probabilistic principal component analysis, and fuzzy stochastic neural network modeling. The advanced signal processing is employed to reduce dimensionality and to address multivariate correlation and data uncertainty for damage prediction. Instead of conventionally using raw time series data, principal components are utilized in the establishment of stochastic neural network model and anomaly prediction. Bayesian interval hypothesis testing metric is then presented to quantitatively compare the predicted and measured data for model validation and anomaly evaluation, thus providing a confidence indicator to judge the model quality and evaluate the equipment status. Bayesian method is developed in this study for denoising the raw data with multiresolution wavelet decomposition, quantifying the model accuracy, and assessing the equipment status. The dynamic stochastic neural network model is established via the nonlinear autoregressive moving average with exogenous inputs approach. It seamlessly integrates the fuzzy clustering and independent Bernoulli random function into radial basis function neural network. A natural gradient method based on Kullback-Leibler distance criterion is employed to maximize the log-likelihood loss function. The effectiveness of the proposed methodology and procedure is demonstrated with the 11-variable time series data and the forced outage event of a real-world centrifugal compressor.

Real-time estimation of hospital discharge using fuzzy radial basis function network and electronic health record data

Hospital resources are scarce and should be properly distributed and justified. Information about how long patients stays in critical intensive care units can provide significant benefits to hospital management resources and optimal admission planning. In this paper, we propose an approach for estimating intensive care unit length of stay using fuzzy radial basis function neural network model. The predictive performance of the model is compared to others using data collected over 13,587 admissions and 54 predictive factors from five critical units with discharges between 2001 and 2012. The proposed model compared to others demonstrated higher accuracy and better estimations. The three most influential factors in predicting length of stay at the early stage of pre-admission were demographic characteristics, admission type, and the first location within the hospital prior to critical unit admission. We have found about 63% of patients with multiple chronic conditions, stayed significantly longer in hospital. Enabling the proposed prediction model in clinical decision support system may serve as reference tools for communicating with patients and hospital managers.

A fuzzy neural network bagging ensemble forecasting model for 72-h forecast of low-temperature chilling injury

On the basis of the daily temperature and precipitation data of Guangxi and the NCEP/NCAR reanalysis data and forecast field data, the paper aims to determine the significant nonlinearity and temporal variability of the forecast quantity series and the overfitting that can easily appear in the forecast modeling of a single fuzzy neural network model and many adjustable parameters that are difficult to determine objectively. Thus, an ensemble forecasting model of fuzzy neural network bagging for 72-h forecast of low-temperature chilling injury is developed. The forecast results of independent samples show that under the same forecast modeling sample (N = 299) and forecasting factor (M = 9), the fuzzy neural network bagging ensemble forecasting model obtains a mean absolute error of 13.91. By contrast, the mean absolute errors of the single fuzzy neural network forecasting model and the linear regression forecast are 15.82 and 18.13, respectively. The fuzzy neural network bagging ensemble forecast error is lower by 12.07 and 23.27%, respectively, compared with the latter two methods, showing a better forecasting skill. This improved performance is mainly due to the ensemble individuals of the fuzzy neural network bagging ensemble forecasting model with playback sampling. Different ensemble individuals are obtained. The ensemble enhances the generalization performance and forecast stability of the fuzzy neural network bagging ensemble forecasting model. Thus, this model has better applicability in forecasting nonlinear low-temperature chilling injury.

Selective Control of a Gas-Turbine Engine

The toolkit of soft computing is based on fuzzy systems, fuzzy neural network models, genetic algorithms, etc., with their strengths and weaknesses. This article analyzes the application of this toolkit to aviation gas-turbine engines (GTEs) used in aircraft and gas-turbine power plants. In the theory of automatic control their GTEs are nonlinear objects without any mathematical description but with one input and several outputs. In the course of running a GTE, it is necessary to continuously control its parameters, such as the combustion chamber temperature, the rotation frequency of the rotor of the fan’s free turbine, and the turbocharger rotor frequency. The GTE is supplied with fuel through a selective regulator that analyzes deviations in the parameters of all of the loops in an asynchronous and logically continuous manner. The multiagent or selector-free control of the GTE is based on designing a fuzzy state controller using linear functions of membership for all of the input parameters in a range from 0 to 1. The interparameter connections are established by fuzzy implication with the help of layers the number of which is equal to the number of terms in a specific term set. The fuzzy composition is made up graphically by the pointwise summation of ordinates of truncated figures according to Mamdani. Defuzzification is performed by the centroid technique, including an even partition along the X-axis of the resulting figure. The fuel consumption of the dispenser allows finding the fan rotation frequency, the turbocharger rotation frequency, and the temperature of the combustion products behind the combustion chamber within specified limits.

Definition of Phishing Sites Based on the Team Model of Fuzzy Neural Networks

This paper solves the problem of defining phishing sites based on building a team model of fuzzy neural networks (FNNs). The main methods of phishing are analyzed. Attention is drawn to the fact that phishing has become widespread on the Internet through the use of phishing sites. The expediency of identifying phishing sites based on the analysis of their URLs is noted. The main approaches to identifying phishing sites are described. The need toimplement an approach based on machine learning by constructing fuzzy neural networks for the creation of fuzzy knowledge bases and their use to identify phishing sites is actualized. Automating the identification of phishing sites based on the neuro-fuzzy approach required solving the problems of collecting and preparing initial data for analysis,building a team model of fuzzy neural networks, and forming a fuzzy knowledge base, as well as conducting research, and assessing the accuracy of identifying phishing sites based on the constructed model. The initial data was formed from various sources. The total amount of initial data was 50,000. Of these, 10 input features for analysis were selected by an expert. After carrying out the correlation analysis, 4 most informative input features were selected for analysis: site lifetime, site rank, URL length, and the registered status of the site. An output feature of the site was its type: phishing or legitimate. After assessing the quality and cleaning the selected data, the resulting sample was formed of 34718 rows, of which 70% were used for learning (24303 rows), and 30% (10415 rows) for testing. A team model offuzzy neural networks was built and a knowledge base was formed on the basis of the data obtained, including 4608 fuzzy rules. Studies have shown that the number of errors of the 1st type in identifying phishing sites is 2.01%, and 2.89% for errors of the 2nd type. The general classification error based on knowledge base rules is 4.9%. The accuracy of identifying phishing sites was 95.1%, which exceeds the accuracy of other classification methods: multilayer neural network, decision tree, linear and logistic regression. The knowledge base formed on the basis of the team model of fuzzy neural networks can be effectively used to identify phishing sites on the Internet.

Delay-Dependent Criteria for Global Exponential Stability of Time-Varying Delayed Fuzzy Inertial Neural Networks

This paper is mainly concerned with global exponential stability of time-varying delayed fuzzy inertial neural networks. Different from previous approaches of variable transformation, we use non-reduced order method. Different from previous non-reduced order method used to investigate the inertial neural networks without time-varying delays, we take the time-varying delayed effects into account. By constructing a modified delay-dependent Lyapunov functional and inequality technique, delay-dependent criteria stated with simple algebraic inequalities are given in order to ensure the global exponential stability for the addressed delayed fuzzy inertial neural network model. The approach applied can provide a new method to study the fuzzy inertial neural networks with time delays via non-reduced order method. Some previous works in the literature are extend and complement. Finally, numerical examples with simulations are presented to make comparisons between the system with delays and without delays, and further demonstrate the validity and originality of the proposed approach.

Design and Realization of the Intelligent Design System for Tunnel Blasting in Mine Based on Database

With the improvement of informatization and standardization of mine development, it is an inevitable trend to apply computer technology to the intelligent design of tunnel blasting excavation. Seeking to the problems existing in the intelligent design system of blasting excavation, a new intelligent design system of blasting excavation is constructed from the perspective of database technology. The new intelligent design system adopts the design of central database system and subuser rights to meet the data synchronization and sharing among different users. At the same time, the new system also extracts blasting design parameters and actual engineering data and designs a more reasonable and comprehensive database structure, database logic, and data word table. Based on the T-S fuzzy neural network model, the intelligent search rules of excavation blasting data are also constructed in the new system. Finally, based on Oracle and VB.NET as the new system development platform, MFC, and ADO as the new system development technology, the new intelligent design system for tunnel blasting and excavation was completed. The implementation of the new system addresses the needs of blasting data information storage and search and lays a foundation for the informatization and standardization of tunnel excavation blasting.

CONSTRUCTION OF INFORMATION TECHNOLOGY FOR DEVELOPMENT OF KNOWLEDGE BASE ON IDENTIFICATION OF URBAN STRUCTURES ON DIGITAL SPACE AND AERIAL PHOTOGRAPHS IN THE URBAN ENVIRONMENT MONITORING

The efficiency and quality of modern city management are directly linked to the relevant monitoring. Improving the efficiency and validity of the obtained data on the urban environment is possible due to the automation of the system of urban objects identification on digital space and aerial photographs, which involves determining their changes. Therefore, a knowledge base (data), which consists of a set of rules, facts, an inference mechanism that can be implemented through the use of deep neural network or hybrid (fuzzy neural network) models, is therefore needed. This academic paper proposes the technology of formal presentation of the information technology construction for the development of a knowledge base on the identification of urban structures on digital space and aerial photographs in automated monitoring of the urban environment. The developed technology identifies an interdependent set of phases, with the possibility of further parallelization of sub-stages, taking into account the peculiarities of knowledge representation (formalization) on the identification of urban structures on digital space and aerial photographs in automated monitoring of the urban environment.

Principal Component Analysis Based Dynamic Fuzzy Neural Network for Internal Corrosion Rate Prediction of Gas Pipelines

Aiming at the characteristics of the nonlinear changes in the internal corrosion rate in gas pipelines, and artificial neural networks easily fall into a local optimum. This paper proposes a model that combines a principal component analysis (PCA) algorithm and a dynamic fuzzy neural network (D-FNN) to address the problems above. The principal component analysis algorithm is used for dimensional reduction and feature extraction, and a dynamic fuzzy neural network model is utilized to perform the prediction. The study implementing the PCA-D-FNN is further accomplished with the corrosion data from a real pipeline, and the results are compared among the artificial neural networks, fuzzy neural networks, and D-FNN models. The results verify the effectiveness of the model and algorithm for inner corrosion rate prediction.

Generalized Dynamic Fuzzy NN Model Based on Multiple Fading Factors SCKF and its Application in Integrated Navigation

Traditional neural network (NN)'s generalization ability is weak, and its prediction accuracy depends heavily on the selection of network structure and training samples, so it cannot be directedly applied to the strapdown inertial navigation system (SINS) and global navigation satellite system (GNSS) integrated navigation system in varied environment. Aiming at these two problems, based on the fuzzy neural network (FNN) model, a new neuron growth-attenuation mechanism is established by introducing the dynamic adjustment idea of network structure. Therefore, the network has a compact structure and good performance, which prevents the network from over-training and over-fitting. Besides, the theory of strong tracking filter (STF) is introduced into the nonlinear filter to design the multiple fading factor square root cubature kalman filter (MSCKF) method for neural network parameter training, which shortens the training time and improves the convergence speed. Results of the simulation and physical experiment verification demonstrate that the generalization ability of proposed model is enhanced and prediction accuracy is improved during the GNSS signal loss. Compared with the pure inertial navigation method, the position errors in latitude and longitude are reduced by 85.00%, 89.71% and the velocity errors in east and north are reduced by 94.57%, 83.11%, respectively by the proposed generalized dynamic fuzzy NN Model Based on MSCKF(MSCKF-GDFNN).

Research on Military Radio Fault Diagnosis Method Based on Fuzzy Neural Network

In view of the complexity, correlation and fuzziness of military radio faults, a fault diagnosis strategy combining fuzzy neural network with expert system is proposed, and the fault diagnosis method based on T-S fuzzy neural network is mainly studied. The fuzzy neural network model was established by MATLAB R2017a, and the simulation experiment was carried out. The experimental results show that the T-S fuzzy neural network model has good generalization ability, fault tolerance ability and self-adaptability, can quickly locate fault causes, greatly improve the diagnostic efficiency, and can be used in fault diagnosis of military radios.

A pretreatment method of wastewater based on artificial intelligence and fuzzy neural network system

A pretreatment method of industrial saline wastewater based on Artificial Intelligence based fuzzy neural network analysis was proposed to improve the pretreatment accuracy of industrial saline wastewater. This method uses a four-layer AI fuzzy neural network model and proposes a graded fuzzy neural network model for pretreatment method of industrial saline wastewater, it includes input layer, fuzzification layer, fuzzy logical layer and output layer, and designs the framework and calculation mode of the fuzzy function block and the neural network module. Finally, the dynamic simulation experiments of dissolved oxygen control in the fifth zone and nitrate nitrogen control in the second zone are carried out based on the simulation benchmark model (BSM1) platform. The experimental results show that this approach can effectively raise the adaptive control accuracy of the system compared with PID, feed forward neural network and conventional recurrent neural network.

Almost automorphic solutions for Clifford-valued neutral-type fuzzy cellular neural networks with leakage delays on time scales

We propose a class of Clifford-valued neutral-type fuzzy cellular neural network models with leakage delays on time scales, give the definition and basic properties of Clifford-valued almost automorphic functions on time scales, and prove a composition theorem of Clifford-valued almost automorphic functions on time scales, and on this basis, we study the existence and global exponential stability of almost automorphic solutions of this class of neural networks. The research in this paper can unify the research of real-valued neural networks, complex-valued neural networks and quaternion-valued neural networks, as well as the research of continuous-time neural networks and discrete-time neural networks. Finally, an example illustrates the feasibility of the results obtained.

Improving a fuzzy neural network for predicting storage usage and calculating customer value

Predicting the behavior of customers plays a crucial role in the quality of resource management and customer services. In this article, a fuzzy neural network model for predicting the customer storage usage is identified. The identified fuzzy neural network is improved and finally the result of the improved fuzzy neural network is compared with some other fuzzy neural network and other prediction methods.

Frequent rule reduction for phishing URL classification using fuzzy deep neural network model

Phishing is an online extensive fraud, which can trick Internet clients into uncovering their mystery data and qualifications, e.g., login id, secret key, charge card number, and so on. Phishing is one of the real computer security dangers looked by the digital world and could prompt monetary losses for both industries and people. In this paper, two datasets (Phishtank and UCI) were considered for malicious URL detection analysis. At first, the optimal features are extracted from the datasets using DBA-based detector module and then seventy-five optimal rules are generated using Association Rule mining based on these features. However, it is having some repetitive features which can lessen the efficiency of the malicious URL detection process. To beat this issue, the Frequent Rule Reduction algorithm along with the classification approach for predicting Phishing Websites should be implemented. The outcomes show that the proposed Fuzzy Deep Neural Network classifier accomplishes the most extreme accuracy in the prediction analysis of phishing, non-phishing, and suspicious URL.

Adaptive synchronization for fuzzy inertial complex-valued neural networks with state-dependent coefficients and mixed delays

In this paper, a class of fuzzy inertial complex-valued neural networks with state-dependent coefficients and mixed delays is considered. We construct the model of fuzzy inertial complex-valued neural networks with state-dependent coefficients and mixed delays for the first time. By dividing the fuzzy inertial complex-valued neural networks into real and imaginary parts, the model is converted into two fuzzy inertial real-valued neural networks. Then, we design a novel adaptive controller, and under the action of this controller, synchronization criteria for the fuzzy inertial complex-valued neural networks with state-dependent coefficients and mixed delays are obtained by Lyapunov's stability theory and the differential equation theory with discontinuous right side. Finally, a numerical example is given to show the correctness of results.

Water Resources Carrying Capacity Assessment in Dali Prefecture Based on T-S Fuzzy Neural Network Model

Water resources carrying capacity (WRCC) assessment is an effective way to evaluate the coordinated development of water resources, economic society and ecological environment, thus, it is significant to study WRCC to realize the harmony among water resources, economy and society in region. In this paper, an index system for evaluating WRCC, which represents the natural attributes and the managerial attributes of water resource, was constructed. Moreover, the T-S Fuzzy Neural Network Model was applied to evaluate WRCC of resource-oriented, management-oriented and comprehensive in Dali Prefecture, respectively. The results showed that evaluation results of WRCC of resource-oriented, management-oriented and comprehensive were different from each other obviously. Evaluation results of comprehensive WRCC accords with the fact. The WRCC in research region was averaged overall. WRCC in Heqing county, Eryuan county, Yangbi county, Dali city, Binchuan county, Xiangyun county and north of Midu county and Weishan county was already in a critical status, but that in the other area of Dali Prefecture was in non-overloading status.

Textual content prediction via fuzzy attention neural network model without predefined knowledge

Text analysis is a popular technique for finding the most significant information from texts including semantic, emotional, and other hidden features, which became a research hotspot in the last few years. Specially, there are some text analysis tasks with judgment reports, such as analyzing the criminal process and predicting prison terms. Traditional researches on text analysis are generally based on special feature selection and ontology model generation or require legal experts to provide external knowledge. All these methods require a lot of time and labor costs. Therefore, in this paper, we use textual data such as judgment reports creatively to perform prison term prediction without external legal knowledge. We propose a framework that combines value-based rules and a fuzzy text to predict the target prison term. The procedure in our framework includes information extraction, term fuzzification, and document vector regression. We carry out experiments with real-world judgment reports and compare our model's performance with those of ten traditional classification and regression models and two deep learning models. The results show that our model achieves competitive results compared with other models as evaluated by the RMSE and R-squared metrics. Finally, we implement a prototype system with a user-friendly GUI that can be used to predict prison terms according to the legal text inputted by the user.