Combination Forecast Model Of Neural Network And Its Application Of House Prices Trend

House price, a hot issue, has been always concerned by the whole society. Because real estate market is a very complicated nonlinear system, a precise description of house price fluctuation is increasingly important to us. At the same time, house price prediction draws more attention from people, so how to choose an appropriate forecasting model to reflect house price is becoming an immediate necessity for us.As to the prediction of many-factored-effected target, nonlinear functional approximation shows us favorable nonlinear mapping capabilities, generalization capabilities and fault tolerance. It is very suitable to apply to the prediction to a complicated multi-factor system. By means of the combination of Neural Network and the weighted geometric average based on the norm of L1, and its effectual application to the analysis of the prices trend of Hefei commodity housing, the paper mainly deal with the following aspects:First, by means of adopting the least-square method and using of EVIEWS software the paper analyzes the statistics of those multi-facets on the house price. Excluding the less influential facet 'completed housing investment' on the target from the seven rudimentary confirm certainties.Second, the paper will establish respective BP neural network, Elman neural network, grey neural network and the support vector machine model from the analyzed data.Third, by virtue of the error of the predicted result and the actual result from single prediction model, the paper sets up the nonlinear combination forecast model based on the weighted geometric average of the norm of L1,solves the model and finds that the weights of the four kinds of single forecast model are 0.32,0.23,0,0.45,which indicates that the error of the grey neural network model is bigger than the other three models. Combination model is insensitive to the grey neural network model. With the use of the established combination model to do prediction and from a competitive analysis of the errors, we will find that the error of the established combination model is obviously smaller than single prediction model's, its evaluation prediction on house price is also better.