Author(s)

Li Qi

Corresponding Author

Li Qi

Abstract

Learning prediction is an important research content of data mining. It is widely used in many fields to describe the potential relationship between prediction indicators and influencing factors. The classical prediction methods have some difficulties in the application of nonlinear system prediction, while the RBF neural network has better nonlinear characteristics, which is especially suitable for highly nonlinear systems, and opens up a new development space for multi-factor time series prediction. In this paper, the prediction model based on RBF neural network is studied in depth, and the dimensionality reduction reconstruction of network input space is studied in detail. RBF neural network was used for modelling training, and the results were compared with BP network. The training speed of RBF network was significantly faster than that of BP network, with better generalization ability. The simulation experiment showed that it could be effectively applied to multi-factor time series prediction.

Keywords

Rbf network; Learning prediction; Neural network; Sequence; Nonlinearity