基于PCA-BPNN算法的房价预测应用研究  

Research on the Application of PCA-BPNN Algorithm in Housing Price Prediction

作　　者：张璐璐 麻晓敏 王星月 孙俊杰 ZHANG Lulu;MA Xiaomin;WANG Xingyue;SUN Junjie(Department of Information Technology,Anhui Vocational College of Grain Engineering,Hefei 230011,China;Department of Basic Sciences,Army Artillery Air Defense Academy,Hefei 230031,China)

机构地区：[1]安徽粮食工程职业学院信息技术系,合肥230011 [2]陆军炮兵防空兵学院基础部,合肥230031

出　　处：《长春工程学院学报（自然科学版）》2024年第2期114-118,共5页Journal of Changchun Institute of Technology：Natural Sciences Edition

基　　金：安徽省职业与成人教育学会教研规划重点课题(azcg44);安徽省高校人文社会科学研究重点项目(2022AH053106);安徽省教育厅高校质量工程项目(2022jpkc041)。

摘　　要：房价是影响人民生活幸福指数的重要因素,因此合理地进行房价预测意义重大。以经典预测数据集——波士顿房价数据集为例,提出一种基于主成分分析(PCA)的3层BP神经网络模型的改进算法PCA-BPNN来进行房价预测。在对数据集进行数据标准化处理和主成分分析降维的基础上,通过调整BP神经网络模型的隐含层神经元数、学习次数等参数来优化预测模型。最后,利用MATLAB对数据进行仿真试验。试验结果表明,提出的模型预测准确率较改进前的BP神经网络模型有所提升,提升幅度最高可达90.4772%。Housing prices is an important factor affecting people′s happiness index,so it is of great significance to predict housing prices reasonably.Taking the classic prediction datasets-the Boston House Price Datasets-as an example,an improved algorithm PCA BPNN based on principal component analysis(PCA)for a 3-layer BP neural network model is proposed for house price prediction.On the basis of data standardization and principal component analysis dimensionality reduction on the datasets,the prediction model is optimized by adjusting parameters such as the number of hidden layer neurons and learning times of the BP neural network model.Finally,it uses MATLAB to conduct simulation experiments on the data.The experimental results show that the proposed model has improved prediction accuracy compared to the original BP neural network model,with a maximum improvement of 90.4772%.

关 键 词：BP神经网络 房价预测 数据预处理 主成分分析 累计贡献率 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]