基于多源数据的房地产价格指数预测模型及实证研究  

Prediction Model and Empirical Study of Real Estate Price Index Based on Multi-source Data

作　　者：朴春慧 武旭晨 蒋学红 李玉红 Piao Chunhui;Wu Xuchen;Jiang Xuehong;Li Yuhong(School of Information Science and Technology,Shijiazhuang Tiedao University,Shjiazhuang 050043,China;Ministry of Information Technology,Bank of China Hebei Branch,Shijiazhuang 050000,China;Department of Housing&Urban Rural Development.Hebei,Shjiazhuang 050051,China;School of Eonomics and Management,Shijiazhuang Tiedao University,Shijia zhuang 050043,China)

机构地区：[1]石家庄铁道大学信息科学与技术学院,河北石家庄050043 [2]中国银行河北省分行信息科技部,河北石家庄050000 [3]河北省住房和城乡建设厅信息中心,河北石家庄050051 [4]石家庄铁道大学经济管理学院,河北石家庄050043

出　　处：《石家庄铁道大学学报（社会科学版）》2020年第2期14-22,共9页Journal of Shijiazhuang Tiedao University（Social Science Edition）

基　　金：河北省住房和城乡建设厅信息中心项目“住房城乡建设行业大数据应用指南”。

摘　　要：房地产的价格变化对社会经济发展有显著的影响,准确预测房地产市场价格变化并对其进行有效调控显得尤为重要,但使用房价作为评估房地产市场的度量指标有一定的局限性。住宅销售价格指数是由国家统计局发布的综合反映住宅商品价格水平总体变化趋势和变化幅度的相对数,为探讨新建商品房住宅销售价格指数的预测方法及其预测有效性,利用与相关的房地产供求关系、社会宏观经济指标、国家货币政策和民众对房价的预期等多源数据,构建了一套房地产价格指标体系。分别使用BP-Adaboost和支持向量回归机两种机器学习算法构建房地产评估模型,同时设计了一个调参算法对支持向量回归机模型进行参数优化。在实证中使用华北某市的房地产月度数据对两种模型进行训练和预测,并与ARIMA模型和经典BP神经网络模型做对比。实验结果表明,BP-Adaboost模型的预测误差最小,使用BPAdaboost模型预测房地产价格指数具有可行性。The price change of real estate has a significant impact on social and economic development.It is particularly important to accurately predict the price change of real estate market and to effectively control it.However,the use of house price as a measurement index to evaluate the real estate market has certain limitations.Residential sales price index is a relative index issued by the State Statistical Bureau,which comprehensively reflects the general trend and range of changes in housing commodity prices.In order to explore the forecasting methods and effectiveness of the new commercial housing sales price index,it utilizes related real estate supply and demand relationship,social macroeconomic indicators,national monetary policy and people’s expectations of housing prices.According to the data,a set of real estate price index system is constructed.Two machine learning algorithms,BP-Adaboost and Support Vector Regression Machine(SVR),are used to construct the real estate evaluation model,and a parameter adjustment algorithm is designed to optimize the parameters of the SVR model.The monthly real estate data of a city in North China are used to train and forecast the two models,and The ARIMA model and the classical BP neural network model are compared with the model proposed in this paper.The experimental results show that the prediction error of BP-Adaboost model is the smallest,and it is feasible to use BP-Adaboost model to predict the real estate price index.

关 键 词：房地产预测 BP-Adaboost算法 支持向量回归机 住宅销售价格指数 

分 类 号：F293.35[经济管理—国民经济]