基于支持向量机回归的房地产批量估价模型研究

发布时间：2018-01-03 15:32

本文关键词：基于支持向量机回归的房地产批量估价模型研究　出处：《暨南大学》2013年硕士论文　论文类型：学位论文

【摘要】：由于目前房地产估价行业发展不够成熟，，相关从业人员素质、估价手段、影响房价的因素等不固定因子会导致估价出现参差不齐的现象。需要一种科学的方案来对估价行业进行规范化管理。本研究在Hedonic估价模型的基础上进行改进，制定出一套合理完善的基于GIS和支持向量机回归的房地产批量估价模型。首先，按不同楼盘、各房屋类型的成交案例对住房自身属性相关数据进行采集，以及通过Google MapsAPI调用GIS系统来批量提取住房所在楼盘周边的区位因素。接着，对不同类型住房的房价进行修正为同样实物状况下的标准房价，按房价对不同价位的楼盘进行聚类，分成低价房、平价房和高价房三大类，并针对每个类型的楼盘进行拆分为训练集和测试集两部分样本；然后，针对各类型住房的训练集样本，分别采用四种核函数以及经验参数法、交叉验证法、遗传算法和基于自适应变异的PSO算法等四类参数寻优方法，得到基于径向基核函数和自适应变异的PSO算法的支持向量机回归方法建立的回归模型估价结果最为精确；并通过对测试集中已知的区位因素进行批量估价，同时也检验了支持向量机回归批量估价模型的精确度，其估价结果明显优于岭回归和BP神经网络这两种估价方法。最后，通过国际估税官协会（IAAO）提供的批量估价效果的评价指标体系对该模型进行综合评价，并为房地产批量估价工作制定了标准化的流程。
[Abstract]:Due to the current development of real estate valuation industry is not mature, the quality of relevant practitioners, valuation means. Factors affecting house prices and other uncertain factors will lead to uneven valuation. A scientific scheme is needed to standardize the management of the valuation industry. This study is based on the Hedonic valuation model. Make improvements. A set of reasonable and perfect real estate batch valuation model based on GIS and SVM regression is developed. Firstly, according to different housing developments, the transaction cases of different types of housing to collect the relevant data of the property of housing itself. And through the Google MapsAPI to call the GIS system to extract the housing estate around the location of factors. The housing prices of different types of housing are revised to the standard housing prices under the same physical conditions, according to the price of different housing prices for clustering, divided into low-price housing, affordable housing and high-priced housing three categories. And for each type of real estate divided into two parts of the training set and test set samples; Then, four kinds of parameter optimization methods, such as kernel function, empirical parameter method, cross-validation method, genetic algorithm and PSO algorithm based on adaptive mutation, are used for the training set samples of each type of housing. Support vector machine regression method based on radial basis function kernel function and adaptive mutation PSO algorithm is obtained. The evaluation results of the regression model are the most accurate. At the same time, the accuracy of the support vector machine regression batch evaluation model is also tested by evaluating the known location factors in the test set. The evaluation results are obviously superior to those of ridge regression and BP neural network. Finally. Through the evaluation index system of the batch valuation effect provided by IAAO( International Association of Appraisers), the model is evaluated synthetically, and a standardized flow chart is established for the real estate batch evaluation.
【学位授予单位】：暨南大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F299.23;F233;F224

【参考文献】