基于人工神经网络的房地产市场预警体系研究

【作者】 王建新；

【导师】 贾生华；

【作者基本信息】 浙江大学 ， 管理科学与工程， 2006， 硕士

【摘要】 随着房地产业在国民经济中支柱产业地位的确定,房地产业的健康发展对国民经济的影响日益显著。科学的房地产市场预警体系能够为市场主体制定决策提供参考,包括政府宏观调控措施的出台,因此有助于房地产业的健康发展。而从房地产市场预警的已有研究来看,由于不能清晰界定市场警兆指标和警情之间的复杂关系,影响了预警的准确性。因此,本文将具有强大的函数拟合能力的人工神经网络方法引入到房地产市场预警体系中,通过拟合出警兆指标和警情之间的映射关系,构建基于人工神经网络的房地产市场预警体系,将有助于提高房地产市场预警的准确性。 在此背景下,本文首先在阅读和整理相关文献的基础上,从经济预警理论和周期波动理论入手,理清了基于人工神经网络的房地产市场预警体系的构建思路。然后,选择国内起步最早、发展最成熟的深圳房地产市场进行实证研究。在导师的指导下,对深圳房地产市场进行了实地调研,对深圳市国土资源与房地产管理局、深圳市房地产估价中心、中山大学岭南学院房地产咨询研究中心、深圳市房地产研究所、招商银行、万科企业股份有限公司、金地股份有限公司、招商局地产股份有限公司等房地产相关企业和部门的专家进行访谈和问卷调查,并收集了1986年以来深圳房地产市场的历年历史数据,获得17个能提供给人工神经网络训练的样本。接着,借助Matlab7.0软件中的人工神经网络工具箱,编写了实现人工神经网络训练的程序,并借助样本对网络进行训练,构建基于人工神经网络的房地产市场预警模型。预警模型的检测结果显示,本文所建立的预警模型具有较好的泛化能力,能胜任房地产市场预警的功能要求。最后,借助所建立的预警模型对2005年深圳房地产市场状况进行了分析,结果显示,2005年深圳房地产市场将处于“正常”状态。 论文的创新点主要体现为两个方面:第一,针对房地产市场预警指标体系概念混乱的现状,从房地产市场景气指标体系入手,构建了房地产市场警情指标体系和警兆指标体系。第二,将人工神经网络方法引入到房地产市场预警中,通过拟合出市场警兆指标和警情之间的复杂关系,构建基于人工神经网络的房地产市场预警体系,提高了预警的准确性。 由于研究经验、能力和时间等方面的限制,本文还存在一些不足。首先,国内房地产市场发展时间较短,可获取样本不多,限制了本文预警体系所能实现的功能;其次,本文预警体系缺乏对单个指标的分析,不利于警源的寻找和排警。为此,后续的研究可在这两个方面加强。更多还原

【Abstract】 As the real estate industry has been the mainstay industry of national economy, it is important to ensure its health development. Scientific early warning system of real estate market is helpful to the healthy development of real estate industry as it can offer reference to the market players, including the macro-control of government. But the current research of early warning of real estate market can not reflect the complex relationship between warning indicators and degree of alarm and loss the veracity of warning. So this thesis introduces the neural network of high function fitting ability into early warning of real estate market, builds a real estate market early warning system based on the artificial neural network by reflecting the complex relation between warning indicators and degree of alarm, and will improve the accuracy of real estate market early warning.Against this background, the thesis gets the idea of building the early warning system based on artificial neural network into shape firstly started with the economy early warning theory and business fluctuation theory. Next, it makes the empirical analysis with a sample of Shenzhen. At the direction of my tutor, interview and questionnaire survey to the experts of institutions such as Shenzhen Municipal Bureau of Land Resources and Housing Management and China Vanke Co.,Ltd were done and the data from 1986 which is related to real estate was collected in Shenzhen. As a result, 17 samples are obtained from these data. On the next step, a program realizing the training of artificial neural network is compiled in the thesis by the neural network box of Matlab7.0, and the early warning model of real estate market is built by training the samples. Testing result indicates that the model built in the thesis is qualified for the early warning of real estate market. At last, the thesis analyzes the Shenzhen real estate market by the early warning model based on the artificial neural network, output of the model indicate that the Shenzhen real estate market of 2005 will on the degree of "normal".The main innovations of the thesis concentrate on two aspects: First of all, to the question of disorder in the concept of real estate early warning indicator, the thesis constructs the warning indicator system and alarm indicator system on the basis of the prosperity indicator system. Second, introduces the neural network into the real estate market early warning and builds a real estate market early warning system based on the artificial neural network and improves the accuracy of real estate market early warning.Being enslaved to research experience, capability and time, there are still some shortcomings in the thesis. Firstly, as the short history of domestic real estate market can not provide enough samples for the train of artificial neural network, capability of real estate market early warning system based on the artificial neural network will be restricted. Secondly, the absence of analyzing in single warning indicator will be not propitious for searching the origin of alarm and obviating the alarm of real estate market. So the next research can strengthen in these aspects.更多还原