Neuro-fuzzy Systems And Neural Financial Prediction A Number Of Issues,

Recently, soft computing technology is a hot point of research. Among the soft computing technology, neural networks and fuzzy system are two most important technologies. Neural networks and fuzzy systems got successful in industrial application. These two kinds of technologies are widely used in automatic control, traffic and flow control in ATM networks, financial prediction, and pattern recognition etc.There are many similarities between neural networks and fuzzy system, for example, function approximation, equivalence between neural networks and fuzzy systems, model free systems etc. There are mainly three kinds of types of the combination of neural networks and fuzzy system: fuzzy operator neural networks, fuzzy information neural networks, and neuro-fuzzy systems.Fuzzy operator neural network, to some extent, is a kind of fuzzification of neural networks. Fuzzy operators replace the arithmetic operators in neural operators of neural networks. Fuzzy information neural networks are used to process fuzzy information. The input and output of fuzzy information neural networks are fuzzy sets. The operation and learning algorithms of fuzzy information neural networks are based on the level cut operation of fuzzy sets and the extension theorem of fuzzy sets. Neuro-fuzzy systems use neural networks as tools to realize fuzzy systems. Fuzzy information neural networks and neuro-fuzzy systems combine the merits of neural networks and fuzzy systems, and offset the demerits of them. Neuro-fuzzy systems and fuzzy information neural networks have the merits not only that the results can be explained by fuzzy rules, the prior knowledge can be utilized by them but have very strong learning capacity. The operation and learning algorithms of neuro-fuzzy systems are simple than that of fuzzy information neural networks. Among these three types of combination, neuro-fuzzy system is the most popular one used in practice.In this paper, after the deed research on fuzzy system identification, the two-phrase structure identification of neuro-fuzzy systems is proposed. And then partition free neuro-fuzzy system and the regularization theory for neuro-fuzzy system are proposed in this paper. In the last, the problems and solutions to neural network-based time series predictor are addressed in this paper. The main contributions of this paper are followed:1 The properties and relations of several identification technologies for...