Design of Credit Risk Evaluation System Based on the Fuzzy Neural Network and Expert System

A new method of ozone forecasting using fuzzy expert and neural network systems

By analyzing the character of oil equipment detection, the intelligent fusion model of detection information of oil equipment has been established. The feature-level fusion algorithm based on fuzzy neural network and expert system has been proposed, in which the expert system has been embedded into fuzzy neural network so that it could choose the membership function and adjust the network structure. The improved PSO algorithm has been adopted to train fuzzy neural network and prune fuzzy rules. Evidence theory has been applied to achieve the decision-making level fusion. Then, the results of feature-level fusion have been taken as the evidences to construct the frame of discernment. On the basis of the generalized evidence combination rule, the conflict evidence combination rule based on the weighted averaging method is proposed, and the prior knowledge in expert system has been utilized to adjust the evidence weights. The research results show that the process of detection information fusion has abilities of adapting and self-learning. This research has significant importance on reliability of improving oil equipment.

After information fusion model has been established, the feature-level fusion algorithm based on fuzzy neural network and expert system is proposed, in which the expert system has been embedded into fuzzy neural network so that it could choose the membership function and adjust the network structure. At the same time, for code tracking loop, two new code phase discriminator algorithms based on DLL structure is proposed. Evidence theory has been applied to achieve the decision-making level fusion. The performances of the two algorithms were studied by using theoretical method and experimental method with analog IF signal data and actual IF signal data respectively. Then, the results of feature-level fusion have been taken as the evidences to construct the frame of discernment. The research results show that the process of information fusion has abilities of adapting and self-learning.

Because the traditional linear vectorization methods have some shortcomings including processing data slowly, being sensitive to noises and being easy to be distorted. Fuzzy rules and its inference mechanism are the assurance of achieving feature fusion. However, the self-learning function of FNN could train its weights; it is difficult to optimize fuzzy rules. Besides, the common FNN training algorithms have low constringency speed and are liable to run into the local optimization.PSO algorithm has high convergence speed and it is simpler on the operation and is more potential on optimizing FNN. Thus, PSO algorithm could be adapted to train FNN weights, and prune the redundancy links, optimize fuzzy rules base. In the paper we present an improving immune genetic algorithm based on chaos theory. The over-spread character and randomness of chaos can be used to initialize population and improve the searching speed, and the initial value sensitivity of chaos can be used to enlarge the searching space. To avoid the local optimization, the algorithm renews population and enhances the diversity of population by using density calculation of immune theory and adjusting new chaos sequence.

Since it is hard to get training set of fuzzy neural network, to understand knowledge rules, and to learn new knowledge through fuzzy expert system, an attribute recognition system based on rough set theory-fuzzy neural network and fuzzy expert system has been put forward. In this paper, it has explained how to use rough set theory to get training set of fuzzy neural network, how to deal with data through fuzzy neural network and fuzzy expert system parallelly, and how to acquire new knowledge from fuzzy neural network to supplement the knowledge database of fuzzy expert system. It has fully utilized the capability of rough set theory that is to simplify large amount of redundant data, the capabilities of fuzzy neural network that are self-learning, fault-tolerant and highly nonlinear mapping, and the capability of fuzzy expert system that is reasoning quality in knowledge. Experiments show the exactness and high-efficient quality of this recognition system, and it has gotten more than 96% correct recognition rate.

Due to the bad performance of the common method for track to track fusion under complicated disturbing environment and radar netting dynamically, an algorithm for track to track fusion based on NFE model (neural network-fuzzy reasoning-expert system) is proposed. By making good use of the self-learning, self-reasoning ability and robustness of fuzzy neural network and expert system, this approach can adapt to the changes of the system environment and find the optimal algorithm. Simulation results show that this algorithm can still get results with high precision even under various complicated conditions such as sensors' reliabilities being low, a large amount of outliers existing in the whole measurement data set, sensors broken down or destroyed, radars netting over again, and so on.

Research of Surveying and Mapping Based on Adaptive Fuzzy Neural Network and Expert System

Proposes a fuzzy neural expert system (FNES) which has a feedforward fuzzy neural network whose input layer consists of fuzzy cell groups and crisp (non-fuzzy) cell groups. Here, the truthfulness of fuzzy information and crisp information of training data is represented by fuzzy cell groups and crisp cell groups, respectively. The expert system has the following two functions: generalization of the information derived from the training data and embodiment of knowledge in the form of the fuzzy neural network; and extraction of fuzzy If-Then rules with linguistic relative importance of each proposition in an antecedent (If-part) from a trained fuzzy neural network. The paper also gives a method to extract automatically fuzzy If-Then rules from the trained neural network. To prove the effectiveness and validity of the proposed fuzzy neural expert system, a fuzzy neural expert system for medical diagnosis has been developed. >

We present an adaptive hierarchical sliding mode control based on fuzzy neural network (AFNNHSMC) for a class of underactuated nonlinear systems. The approach is applied to the problem of high-precision trajectory tracking. The underactuated nonlinear system is viewed as several subsystems. One subsystem is used to design the first layer sliding surface, which constructs the second layer sliding surface with another subsystem. When the top layer, the nth layer, includes all the subsystems, the design process is finished. Meanwhile, the equivalent control law and the switching control law are achieved at every layer. Because the hierarchical sliding mode control (HSMC) law relies excessively on the requirement of detailed information of the underactuated dynamic system, and because that method causes an inevitable chattering phenomenon, an online fuzzy neural network (FNN) system is applied to mimic the HSMC law. Moreover, the bounds of system uncertainties, time-varying external disturbances, and modeling error caused by the fuzzy neural network system are estimated online by a robust term. The stability of the closed-loop system is guaranteed based on the Lyapunov theory and the Barbalat's Lemma. Finally, the example of a single-pendulum-type overhead crane system is simulated and used to verify the effectiveness and robustness of the proposed method compared with the conventional HSMC method.

Summary form only given. Cooperative schemes between neural net models and fuzzy expert systems were investigated. Performance of a specific neural net model with back-propagation models is discussed. Information is initially received by the expert system (ES) in a form that is useful for certain tasks. Information is represented in more than one form, reflecting interpretations of uncertainties and ambiguities in data instances. The multiple representations emerge from different scoring methods applied at input time to designated uncertainties, and on subsequent analysis, through use of linguistic relative preferences choices. The ES responds to new data as it arises, so that its adjustments do not required massive relearning. The information contained in the fuzzy system is then made available to a modified form of interactive activation and competition model (MIAC). MIAC is implemented as a collection of objects in an object-oriented system. >

Engine has a high chance of failure, it usually accounts for about 40% of vehicle failures. Study expert system of engine fault diagnosises that it can locate fault timely and accurately, and enhance efficiency. However, the traditional expert system has shortcomings so as inefficient inference and poor self-learning capability. The fuzzy logic and traditional neural networks are combined to form fuzzy neural networks, they are established a model of fuzzy neural network (FNN) of fault diagnosis, and that the model is applied to engine fault diagnosis, complementary advantages, to effectively enhance efficiency of inference and self-learning ability, its performance is higher than the traditional BP network.

The paper describes the different classifications of expert systems. There are mainly five types of expert systems. They are rule based expert system, frame based expert system, fuzzy expert system, neural expert system and neuro-fuzzy expert system. We discussed the expert systems based on their knowledge representation, inference engine, working of the system and user interface. At the end, we provided comparative study of above five types of expert systems based on different parameters. We can say that choice of an expert system depends on the domain requirements.

As you may know from recent e-mail, BUSEFAL Vo1.76, Obituary, and Fuzzy Sets and Systems (FSS), Vo1.104, No.2, Obituary, Prof. Ernest Czogala passed away on October 8, 1998. First, I would like to express my sincere condolences to his eldest daughter Theresa Czogala-Koczy and son. The call for papers for this special issue has been answered from all over the globe. This issue includes the first seven accepted. The next issue will include those remaining. Since a formal obituary was provided by Professor Witold Pedrycz, a student of Ernest, in FSS as indicated above, I will add a few words of my own here. I first met Prof. Czogala when I was a visiting professor at the University of Alabama at Birmingham (UAB) Department of Computer and Information Sciences in 1990-1991. He was visiting the Mathematics Department, working with Professor James J. Buckley. When I first met Ernest, I intuitively felt that he, Jim (Prof. Buckley), and I would accomplish outstanding work in the near future. I promised to invite Ernest and Jim to Ibaraki University Department of Computer and Information Sciences in summer 1991. After my sabbatical at UAB, I received a letter from Ernest, asking, "Could I really visit your university for a month?" I immediately sent a formal invitation letter to Ernest and Jim. Ernest purchased an airline ticket between Poland and Japan - a princely sum equal to 6 months of his salary! He arrived August 19, 1991, at Narita Airport around 07:30 aboard All Nippon Airways from Poland via Wien. I drove a Toyota Camry to Narita from Hitachi City, Ibaraki Prefecture - a 2.5-hour excited jaunt. My wife Madoka and I went to pick him up. He appeared easygoing, shod in flip-flops. Madoka and I took him to Narita-san Temple - and found he had only 50 DM with him! After spending a few hours at Narita-san and eating lunch, Jim arrived from Birmingham. We picked him up and returned to Hitachi. Ibaraki University has a very old-fashioned guesthouse - but it only cost 700 a day. The next day, a coup d'etat shook Moscow. Ernest applauded, saying "Poland will be free." From that day, we started our cooperative research in an un-air-conditioned office at Ibaraki University Department of Computer and Information Sciences. Outside, it was over 35°C, which made it 38°C in the office. During his stay, we wrote many papers. One recently appeared carefully selected based on the Science Citation Index, "On the Equivalence of Neural Nets and Fuzzy Expert Systems," FSS, Vol. 100 Supplement, pp. 145-150 (1999). This short proof paper was outstanding. The original paper was submitted to Professor H.-J. Zimmermann, Founding Editor of FSS, in 1991. The paper was soon revised, accepted for publication in 1992, and published January 25, 1993. We have published many journals and refereed International Conference papers. I recollect papers worked on with Ernest. They include: 1. Hayashi, Y, Buckley, J.J. and Czogala, E., Systems Engineering Applications of Fuzzy Neural Networks, Journal of Systems Engineering, Vol.2, pp.232-236 (1992). 2. Buckley, J.J., Hayashi, Y and Czogala, E., On the Equivalence of Neural Nets and Fuzzy Expert Systems, Fuzzy Sets and Systems, Vo1.53, No.2, pp.129-134 (1993). 3. Hayashi, Y, Buckley, J.J. and Czogala, E., Fuzzy Neural Network with Fuzzy Signals and Weights, International Journal of Intelligent Systems, Vol.8, pp.527-537 (1993). In 1995, Ernest came to Yokohama, Japan, to present his paper at FUZZIEEE/IFES'95. Madoka, my 2-year-old son Yuichiro and I met him and ate Chinese lunch together. We promised to meet again. In 1997, I met him again at Barcelona for FUZZ-IEEE'97. He asked about the amount of grants I had applied for. I replied 700 million for three years. "That's beyond all belief!" Unfortunately, my request was not met... Ernest, Jim, and I would have liked to have a delicious beer commemorating the FSS Vo1.100 Supplement. But although we never got to share that beer, the papers - the fruit of great efforts - never died. I believe our publication will continue to be referenced by many researchers all over the world. The FSS paper also symbolizes the position of associate professor for me at Ibaraki University. I was 34 years old at the time. Those of you wanting to get in touch with Ernest's oldest daughter are asked to send inquiries to the following address: Theresa Czogala-Koczy ul. Kochanowskiego 29-15 44-100 Gliwice, Poland We thank Drs. Toshio Fukuda and Kaoru Hirota, Editors in Chief of the JACI, for accepting my proposal for this special issue. Special thanks also go to the referees for their kind cooperation, devotion, and rigorous review! We also thank Mr. Yasushi Inoue for his excellent management of the editorial work.

Recently, fuzzy set theory and neural networks system are advanced in many engineering field. The automatic reservoir operation for flood control is generally designed to lighten a troublesome workload for the administrator of dam in Japan. Therefore, in this study, the author’s apply two systems of fuzzy and neural networks to the reservoir operation for flood control. The author’s consider to put the control rules of reservoir operation, the information obtained by inquires to actual reservoir operator and the hydrological characteristics in the basin, into the reservoir operation of the dam supporting system. This system of reservoir operation for flood control is that neural networks is applied to the decision of the operational line and fuzzy set theory is applied to the decision of operational volume, that is, release discharge from reservoir of dam. It is obvious that application of reservoir operation gate for flood control by the use of both fuzzy set theory and neural networks system is effective. Transactions on Information and Communications Technologies vol 19, © 1997 WIT Press, www.witpress.com, ISSN 1743-3517

One of the requirements for the operation and planning activities of an electrical utility is the prediction of load for the next hour to several days out, known as short term load forecasting (STLF). This paper presents the STLF of the Iranian national power system (INPS) using artificial neural networks (ANN) and fuzzy expert systems (FES). The ANN is trained with the load patterns corresponding to the forecasting hours and the forecasted load is obtained. The FES modifies the initial forecasted load for the special holidays and also in the case sudden changes in temperature. A data analyser and a temperature forecaster are also included in the NRI STLF (NSTLF) package. The program has satisfactory results for one hour up to a week prediction of INPS load.