Soft computing(SC) is an evolving collection of methodologies, which aims to exploit tolerance for imprecision uncertainty, and partial truth to achieve robustness, tractability, and low cost. SC provides attractive opportunity to represent the ambiguity in human thinking with real life uncertainty. Fuzzy logic (FL) Neural Networks (NN), and Evolutionary Computation (EC) are the core methodologies of soft computing. However, FL, NN, and EC should not be viewed as competing with each other ,but synergistic and complementary , instead. SC is actually the combination or fusion of each methodology which yields new computational capabilities (hybrid systems). Soft computing is causing a paradigm shift (breakthrough) in engineering and science fields since it can solve problems that have not been able to solved by traditional analytic methods (Tractability (TR)). In addition, SC yields rich knowledge representation ( symbols and patterns). ,flexible knowledge acquisition( by machine learning from data and by interviewing experts), and flexible knowledge processing ( inference by interfacing between symbolic and pattern Knowledge), which enable intelligent systems to be constructed at low cost (high machine intelligence quotient (HMIQ)): cognitive distributed artificial intelligence.

Later, chaos computing and immune networks were added to explain so-called complex systems. This borrows ideas from biology , is so-called reactive distributed artificial intelligence.

This plenary talk starts with neuro- fuzzy hybrid systems (FN) for time series analysis and prediction. By taking advantages of fuzzy systems and neural networks, a fast and accurate Sugeno’s type-I fuzzy system(Type-I fuzzy system) is implemented with the combination of the Gaussian radial basis function network(GP-RBFN) and the time delayed neural network (TDNN), which is based on local modeling using fast general parameter(GP) learning and adaptive algorithms. The proposed GP algorithm applied to adaptation and learning for neural networksis very suitable to parameter optimization. of such local linear models in blended multiple model structure. It is applied to a fault detection application. It is experimentally confirmed that the developed fuzzy neural network is more accurate and faster than the RBFN . Then it is followed by FN + chaos computing, FN +fractal computing, FN + wavelet, FN + immune network, GMDH + genetic programming. Lastly, data mining techniques for nonlinear time series analysis and prediction are discussed. . on the basis of Zadeh’s proposal: i.e., "From Manipulation of Measurements to Manipulation of Perceptions-Computations with Words", that is a data mining technology, knowledge easily comprehensible by humans is extracted by obtaining the features of the time series using a Hurst exponent ,a fractal analysis method, , and an autocorrelation analysis method. In order to extract the knowledge, decision making rules comprehensible by humans using the features are derived with rough set theory. Finally the knowledge is embedded into the structure of the Time Delayed Neural Network (TDNN). The excellent prediction accuracy is obtained.