Research Of Commodity Housing Price Forecasting Based On RS-SVM

Basic necessities of life is the most basic human needs in the country, people have solved the basic food and clothing under the condition that the housing problem has increasingly become the focus of social problems, the past decade rising prices triggered a large number of speculative buyers and panic buyers, in order to better grasp the trend of prices, so that consumers rational purchase, at the same time to the real real estate developers investment decision and the government to formulate policies to provide a reference, based on the summary of the existing price prediction methods, aiming at the shortcomings of the existing methods, this paper proposes a support vector machine with rough set price prediction method,the prediction model is the first rough set theory on the primary effect of housing price index reduction based on the index, and then after the reduction of the system based on support vector machine regression prediction on the price,.The main contents of this paper is introduce the related concept of real estate and housing, and on the basis of previous research summarized four categories of factors affecting prices, and put forward the principle of index selection; summarizes several commonly used method to predict prices, considering the shortcomings of existing methods, put forward RS-SVM prices in addition, prediction model, introduces the basic theory of rough set and support vector machine, and discusses the existence of complementarities between these two theories; constructs the flow chart of the RS-SVM model, and a detailed description of the implementation steps of the prediction model; taking Ningbo city as an example, according to the principle of determining index from four major factors affecting the prices of selected 30 indicators, using Rosetta software for the reduction of the index attribute, select the first 18 from the obtained all minimum condition index attribute set Current number of most indexes which influence index system of Ningbo City real, followed by libsvm toolbox epsilon-SVR regression function from 2002 to 2012 a total of 11 years of data as training samples, using improved grid seeking optimum method and cross validation method to find all root mean square error of minimum of a set of reference value and to the set of parameters for 11 years ago prices fitting and in 2013 and 2014 after two years of house prices forecast, and the predicted results and the BP neural network to predict the results of comparison. The results showed that the RS-SVM model than BPNN can have higher prediction accuracy.