Forecasting and identification of stock market based on modified RBF neural network

A financial index forecasting model based on modified RBF neural network is proposed to find important points of stock index which can solve market identification problem. K-means algorithm is used to search initial center parameters of neurons and adjust optimal structure of network. And gradient descent method is set to search optimal centers through intelligent learning the operating mode of stock market which can overcome random design of network parameters. The forecasting index system of model is set which involves Shanghai Composite Index´s price and volume and selection strategy of sample time range was proposed to study a full cycle of stock market rules. It can improve precision and stability to map nonlinear function by the proposed model in Shanghai Composite Index forecasting, compared with other neural network models. Pressure levels of stock market determined by modified RBF model can support stock investment decision.