基于神经网络与支持向量机的组合预测模型——以房地产价格预测为例  

COMBINATION FORECAST MODEL BASE ON NEURAL NETWORK AND SVM ——Its Application of House Prices Forecast As an Example

作　　者：徐富强[1] 谢如龙[1] 郝江锋[1] 

机构地区：[1]巢湖学院数学系,安徽巢湖238000

出　　处：《巢湖学院学报》2012年第6期1-7,共7页Journal of Chaohu University

基　　金：巢湖学院自然科学研究资助项目(项目编号:XLY-201101;XLY-201102);安徽省高校省级科学研究项目(项目编号:KJ2012B116)

摘　　要：预测问题一直是科学与管理领域研究的热点,备受广大科技工作者的关注。为了解决单一预测模型在预测问题上的缺点和误差控制问题,本文提出一种组合预测模型———以误差绝对值和达到最小为准则的线性组合预测模型。这种模型以误差绝对值和达到最小为依据,计算出每个单一模型在预测结果中所占的权重,建立组合预测模型。该组合预测模型较好的利用了神经网络与支持向量机在多影响因子下训练的优势,进一步提高预测精度。将其应用于合肥市房地产价格预测,可以达到理想效果。Forecasting problems,a hot issue in fields of science and management,has been always concerned by scientists and technologists. In order to solve the disadvantages of the single forecast model and error control＂ s problem, this paper presents a combination forecast model -- Error sum of absolute values of minimum criteria for the linear combination forecas model This model take the minimum error absolute values as the basis, calculate for each single model＂ s weight in predicting outcome.Then,establish a combination forecasting model. The model has better advantage by combination of neural network and support vector machine in multiple factors training, further improve the accuracy of prediction. Its application in Hefei City house price forecast, achieve the desired effect.

关 键 词：BP神经网络 ELMAN神经网络 灰色神经网络 支持向量机 组合预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]