Time-Series Forecast Based On Bayes Network

Forecast is absolutely and necessarily important link and premise before decision-making and planning. Time series forecast is an important research aspect in forecast field, it is widely used in weather, astronomy, electric power, medicine, biology, economy, finance, computer etc. The Bayes Network (BN) proposed by Pearl, also called belief network or probability network, is a uncertainty model for knowledge representation based on probability theory and graph theory. It's distinct performance in representing and reasoning about uncertainty makes it becoming one of a hot research topic in artificial intelligence. The combination between Bayes network and Time series forecast forms into a new modeling and prediction method. In this article, we used Bayes network in Time series forecast field, put forward and established three forecasting model based on Bayes network ,which are static Bayes network forecasting model, static N-Bayes network forecasting model and the classification static Bayes network prediction model.The purpose of this paper is to investigate the time series forecast model based on Bayes network. In this article,first, we introduced the theories of Bayes network, next we presented basic concept, principle, consequence and study of Bayes network, and take the condition probability as priority. Second, we introduced the theories of time series forecasting, and take the Autoregressive Integrated Moving Average Models (ARIMA) as priority to explain time series forecasting model and forecast method in detail. Third, By comprehensive use the theory of Bayes network and time series forecast, we put forward three kinds of forecast models, which are static Bayes network forecasting model, static N-Bayes network forecasting model and the classification static Bayes network prediction model. The three forecast models are put forward by layer upon layer. static Bayes network forecasting model is the simplest model and also the most primitive forecasting model .static N-Bayes network forecasting model is the model which increasing the knot number and network complicated degree on the former, the classification static Bayes network prediction model is based on the data classification and use the Bayes network in static Bayes network forecasting model to forecast. Finally, we applying three kind Bayes network forecast models to an experiment and reaching the correlation conclusion.This paper has summed up predecessor's research result in Bayes network and time series forecast, some self viewpoints based on them are presented and also put it into practice. By the experiment, we know the Bayes network is able to modeling and forecast for time series. So we believe the application of Bayes network in time series forecast field will have a broad prospect.