| 引用本文: | 唐朝君, 章 平,包象琳,徐晓峰.GAN-BPM:基于 GAN 的子市场划分房屋定价模型(J/M/D/N,J:杂志，M：书，D：论文，N：报纸).期刊名称,2025，42（6）：105-114 |
| CHEN X. Adap tive slidingmode contr ol for discrete2ti me multi2inputmulti2 out put systems[ J ]. Aut omatica, 2006, 42(6): 4272-435 |
|
| 摘要: |
| 目的 针对城市房屋市场数据中存在着分布和类别不平衡以及特征价格模型(Hedonic Price Model,HPM)面
对房价与特征关系可解释性不足的问题,提出一种基于对抗生成网络(Generative Adversarial Net,GAN)的子市场
划分房屋定价模型。 方法 在改进子市场划分房屋定价模型时,首先引入 GAN 作为数据增强技术模块,生成具有多
样性和逼真度的合成样本,增加样本的多样性,提高模型的泛化能力;接着将 GAN 数据增强和子市场划分的房屋
定价模型相结合;最后,依据房屋所属子市场的概率预测房价,并分析对房价的关键影响因素,提升预测精度和可
解释性。 结果 将模型与 5 个现有模型以及未加入 GAN 的子市场划分房屋定价模型,从平均绝对百分比误差、平均
绝对误差和均方根误差 3 个方面进行对比;通过使用杭州市 2020 年的房产数据,对模型的算法性能进行测试。 结
论 实验证明引入 GAN 数据增强技术模块后,模型在房地产价格预测方面优于其他对比模型,并且具有可解释性
的优点。 |
| 关键词: 房价评估 GAN 数据增强 房产数据 位置特征 子市场 |
| DOI： |
| 分类号: |
| 基金项目: |
|
| GAN-BPM Sub-market Division Housing Pricing Model Based on GAN |
|
TANG Chaojun ZHANG Ping BAO Xianglin XU Xiaofeng
|
|
School of Computer and Information Anhui Polytechnic University Wuhu 241000 Anhui China
|
| Abstract: |
| Objective Addressing issues of distribution and class imbalance in urban housing market data along with
insufficient interpretability of the Hedonic Price Model HPM in explaining the relationship between house price and
features a housing pricing model for sub-market division based on generative adversarial networks GAN is proposed.
Methods In improving the sub-market division housing pricing model GAN is first introduced as a technical module for
data enhancement to generate synthetic samples with diversity and fidelity to increase the diversity of the samples and
improve the generalization ability of the model. Subsequently GAN-based data augmentation is integrated with the submarket division to develop a housing pricing model. Finally housing prices are predicted based on the probability of the
submarkets to which the houses belong and key influencing factors on the house prices are analyzed to improve the
prediction accuracy and interpretability. Results The proposed model is compared with five existing models and a submarket division housing pricing model without GAN based on average absolute percentage error average absolute errorand root mean square error. Algorithm performance is tested using 2020 real estate data from Hangzhou.
Conclusion Experimental results demonstrate that after introducing the data augmentation technology module based on
GAN this model is superior to comparative models in predicting real estate prices and has the advantage of
interpretability |
| Key words: housing price evaluation GAN data augmentation real estate data location features sub-market |
