Universal Fuzzy Inference Systems-Theory And Application

Computational Intelligence(CI) actually uses the bionics ideas for reference, which origins from emulating intelligent phenomenon in nature and attempts to simulate and reappearance the characters of intelligence such as learning and adaptation, so that it becomes a new research domain for constructing the nature and engineering. CI is very fit for solving the difficult and inextricable problems which can't be modeled by traditional technology and is also defined as a methodology involving some algorithms that exhibit an ability to learn and/or deal with new situations. Among them, fuzzy logic aims at modeling the imprecise modes of reasoning and simulates human ability to make rational decisions in an environment of uncertainty and imprecision, which has great advantages in solving both quantitative or qualitative problems and human cognitive problems such as reasoning, evaluating and decision-making.From comprehensive collecting and understanding definitions of Computational Intelligence, this paper introduces the SMB (Simulation-Mechanism-Based) classification method for CI according to the computational mechanisms of CI branches, which begins by reviewing the essence of CI and existing classification methods. SMB method divides all branches into three categories:organic mechanism simulation class(OMS), inorganic mechanism simulation class(IMS) and artificial mechanism simulation class(AMS), through further study on which, we conclude all the branches in organic mechanism simulation class and summarize the general computational model for each OMS sub-class. SMB aims at further discussion on the essence of all branches in these three classes and sub-classes, which offers an academic guide for further study on CI hybrid methods.For solving the major shortcomings of traditional fuzzy inference system(FIS), this paper presents a universal fuzzy inference system (UFIS) based on full study of fuzzy inference system computing mechanism and the nature of fuzzy logic, the basic idea of which is to find the universal operators applied to fuzzy reasoning thus break through inherent limitations of traditional inference operators, and consider the importance(weight) of each input and each rule during the reasoning process. This model has advantages in universal expression of fuzzy inference operator and importance factor of each input and each rule, which is trying to establish fuzzy inference system that can fully reflect the essence of fuzzy logic and human thinking pattern. Through research on the universal nature of fuzzy inference operators, this paper presents a kind of universal aggregation operator "Agg" from understanding the computing essence of existing aggregation operators, and therefore concludes the universal aggregation theory for the multi-object (criteria) decision-making problems.In order to ensure adaptive capacity of UFIS, from the comparative analysis of the existing variety of fuzzy neural network models and discussion on functional equivalence and complementarities between neural networks and fuzzy logic, we obtain the Adaptive Universal Fuzzy Inference System(A-UFIS) by combining UFIS with a feed forward-type neural network according to the basic principle and structure of adaptive fuzzy inference system. A-UFIS model structure and its parameter updating formula are discussed in detail.For verifying the validity of A-UFIS presented in this paper, we apply its special cases. ANFIS, M-AUFIS and Agg-AUFIS into the traffic level of service(LOS) evaluating problem and establish the corresponding level of service (LOS) evaluation models. The experiment results show that:A-UFIS models have super approximating capacity and its essence is a kind of universal approximator; UFIS models after training have great function of non-linear mapping which can be used in non-linear system (complex system) modeling, analyzing and forecasting. A-UFIS as an effective hybrid algorithm in Computational Intelligence provides a new approach and theoretical support for us to solve the complex problemsComputational Intelligence is the data-based intelligence. It is the advanced stage in intelligence theory development and is bound to have brilliant prospects in the future.