基于最优加权法的组合预测模型在海口市房价预测中的应用  

Application of Combined Forecasting Model Based on Optimal Weighting Method in Forecast of Housing Price in Haikou City

作　　者：陈嘉彤 左剑凯 陈铖 Jiatong Chen;Jiankai Zuo;Cheng Chen(Aviation Engine Academy, Shenyang Aerospace University, Shenyang Liaoning;School of Computer Science and Technology, Shenyang Aerospace University, Shenyang Liaoning)

机构地区：[1]沈阳航空航天大学航空发动机学院,沈阳辽宁 [2]沈阳航空航天大学计算机学院,沈阳辽宁

出　　处：《统计学与应用》2018年第6期569-579,共11页Statistical and Application

摘　　要：针对房价预测问题,建立了基于最优加权法的组合预测模型对房价进行预测。选取多个主要影响房价的指标和历史信息两个方面分析,分别建立BP神经网络模型和NAR神经网络模型对房价进行预测,并通过试验法确定网络的结构。采用最优加权法,建立以组合预测模型的误差平方和为目标函数的非线性规划模型,确定了两种模型对应的权值。以海口市2007~2017年的房价及其影响因素数据为基础,对三种模型进行仿真,检验结果表明,组合预测模型的预测误差小于单一模型,比单一模型的误差更稳定。并由文中建立的组合预测模型,给出海口市未来五年的房价预测。Aiming at the problem of house priceforecasting, a combined forecasting model based on the optimal weighting methodwas established to forecast the house price. The analysis of two major indicatorsaffecting housing prices and historical information was carried out. BP neuralnetwork model and NAR neural network model were established to predict housingprices and the structure of the network was determined by experimental methods.The optimal weighting method is used to establish a nonlinear programming modelwith the sum of squared errors of the combined forecasting model as the objectivefunction, and the weights corresponding to the two models are determined. Basedon the data of housing prices and its influencing factors in Haikou City from2007 to 2017, the three models are simulated. The test results show that theprediction error of the combined forecasting model is smaller than the singlemodel and more stable than the single model. And the combined forecasting modelestablished in the paper gives the housing price forecast for Haikou in thenext five years.

关 键 词：房价预测 组合预测 NAR神经网络 BP神经网络 灰色预测 

分 类 号：F2[经济管理—国民经济]