Statistical Prediction Based On Prior Information And Its Application

Regression is a mathematical statistic method which is to study the relationship between variables and variables. It uses regression analysis based on one or several values of independent variables to predict dependent variable values.Bayesian linear regression model, it differs from traditional forecasting methods because of its use of prior information from experience and historical data. It's a time series forecasting method for the study of dynamic model. The prior distribution reflects the understanding of the overall parameters of the distribution before the test, after access to the sample information, the understanding has changed, and the result is reflected in the posterior distribution. That is to say the posterior distribution combines the prior distribution and sample information. The goal of Bayesian statistics is to learn from experience, put the historical information and sample likelihood function together. It is being more and more widely used in the statistical prediction model.This article summarizes the basic ideas of Bayesian statistical methods, given the basic ideas of selection of the prior distribution, parameter estimation and hypothesis testing, discusses the basic theory and its characteristics of linear model based on Bayesian. This article studied the linear and multiple linear regression models which based on Bayesian methods, and established a dynamic linear prediction model based on conjugate prior distribution. The dynamic prediction model was applied to forecast the water level of the Three Gorges Project closure and electricity consumption of someone province, and the prediction results are satisfactory, which indicate that the model has certain advantages.