City Residential Real Estate Sales Price Index For Grading Forecast Method Research

In recent years, the real estate market of our country has been developing rapidly, the ups and downs of house price has gradually become the concern of the society, especially in some first-tier cities, the house price has been speeding up much faster than expected, although the country have been carried out many policy to regulate the real estate market, the house price is still increasing rapidly, which has been beyond the vast majority and the risk problem of the real estate industry has been increasingly outstanding. To guide the government, citizens and developers correctly and to research the law of development of real estate, it’s of significant meaning to predict the price of real estate with scientific methods.Based on the real estate housing sales price index as the research object, according to the related theories of real estate price, this paper introduces the composition and characteristic of real estate prices, and analyze factors that affect China’s real estate prices from different aspects. As an indicator of the real estate market, real estate price index can show the rule of the real estate market changes in a more reasonable way, choosing different real estate price index can reflect the development of the real estate market from different angles, introduce the real estate price index of establishment method systematically, at the same time, select the median method as the method of real estate price index, and develop the days and weeks price index of two time scales by using data collected from real estate residential sales price. Chaos theory believes that the real estate price index in the single time series contains many information affecting it’s changing factors, the phase space characteristics of the real estate price index will be restored as long as reconstructing time series, avoiding the influence of subjective factors. Based on this, this paper introduces the wavelet neural network and support vector machine(SVM) model for time series prediction. First of all, decide the chaos characteristics of day level price index and weeks level price index respectively, in the meantime, using the method of mutual information and Cao’s method to calculate the delay time and embedding dimension, and undertake phase space reconstruction for day level price index and weeks level price index. Then, by using phase space reconstruction based on wavelet neural network and PSO-LSSVM respectively, establish classification prediction model to forecast of days level price index and weeks level price index, through the example analysis, PSO-LSSVM based on phase space reconstruction model is more suitable for the prediction of short-term real estate price index, which provides a new train of thought for the real estate price index prediction research.