Prediction Method And Predictive Study Of Chinese Major Urban Real Estate Price Index

This paper aims to study how to establish a comprehensive system of forecasting price index time series, in order to provide prediction index change, revealing the inherent characteristics of the attribute index of time series and the relationship with the external environment. Through the analysis of feature attribute index time series, that is smooth or not smooth, linear or nonlinear, with or without effect, gets the conclusion that in the index time series modeling and forecasting system, as long as there are three kinds of models, namely ARMAmodel, GARCH model and nonlinear time series model. In this paper, nonlinear time series model with nonlinear real function approximation ability of multilayer radial basis function( RBF) network, so that the model has high modeling and forecasting precision.This paper with the system to modeling and forecasting year and the chain price index time series of 70 major cities in China, and got a lot of valuable conclusions.One is the most index time series is stable, little unstable are the first difference stationary, the time series has its intrinsic change rule. Secondly, the price index time series is predictable, but the prediction accuracy is uneven, there are high and low.Furthermore, part of the price index time series has the ARCH effect of this time series, if it is linear, then use the GARCH model to predict with high accuracy, if the non-linear, use the multi-layer radial basis function network prediction to predict and forecasting accuracy can have a corresponding increase.