| 摘 要: 为合理征收房产交易契税,需要对房产做出准确估价。鉴于目前的房产估价研究大多使用传统评估方法,为提升房产估价效果,本研究采用机器学习方法,基于房屋区位因素、邻里因素和建筑因素设计估价指标体系,使用某市存量房交易数据,构建Random Forest模型、LightGBM 模型和XGBoost模型进行房产交易估价对比。实验结果表明,XGBoost模型的表现最佳。为进一步提升模型性能,采用贝叶斯优化对XGBoost模型的参数组合进行调整。改进后的模型的MAE指数提高了0.4百分点,MAPE指数提高了1百分点,RMSE指数提高了0.6百分点,有效提升了房产估价的准确性,也为XGBoost算法在房产交易领域的应用积累了实证经验。 |
| 关键词: 房产估价;XGBoost;估价模型;存量房 |
|
中图分类号: TP391
文献标识码: A
|
|
| Research on Real Estate Valuation Model Based on Machine Learning |
|
LIU Yue1, LIU Congjun1,2
|
(1.School of Computer Science, Jiangsu University of Science and Technology, Zhenjiang 212000, China;
2.Jangsu KeDa Hui f eng Technology Co., Ltd., Zhenjiang 212000, China)
18864682521@163.com; liu_cj@163.com
|
| Abstract: To reasonably collect property transaction deed tax, it is essential to accurately value real estate. Given that most current real estate valuation research relies on traditional assessment methods, this study adopts machine learning techniques to improve the effectiveness of property valuation. An indicator system for valuation is designed based on factors such as housing location, neighborhood characteristics, and building attributes. Utilizing transaction data from existing houses in a certain city, Random Forest, LightGBM, and XGBoost models are constructed for comparing real estate transaction valuation. Experimental results indicate that the XGBoost model performs the best. To further enhance the model's performance, Bayesian optimization is employed to adjust the parameter combinations of the XGBoost model. The improved model shows an increase of 0.4 percentage points in the MAE index, a 1 percentage points enhancement in the MAPE index, and a 0.6 percentage points increase in the RMSE index, effectively boosting the accuracy of property valuation and providing empirical experience for the application of the XGBoost algorithm in real estate transactions. |
| Keywords: real estate valuation; XGBoost; valuation model; existing houses |
