基于Stacking集成学习的Lasso-GBDT组合房价预测模型研究

Research on Lasso GBDT Combined House Price Prediction Model Based on Stacking Ensemble Learning

房地产行业是我国国民经济的重要组成部分,关乎国计民生。房价预测的准确性与稳定性,对政府、开发商和广大市民均具有重要的现实意义。本文利用网络爬虫方法获取银川市2015年4月至2021年3月安居客房产信息服务平台样本住宅价格数据,分别利用Lasso模型和梯度提升决策树(GBDT)模型对银川市房价进行预测;综合两种预测方法的优点,建构基于Stacking集成学习的Lasso-GBDT组合回归预测模型。通过实例预测结果比较,组合预测模型预测精度均在0.98以上,能有效避免病态数据对拟合程度的影响,较单项预测模型有更高的准确性和稳定性。

The real estate industry is an important component of China's national economy,which is related to the national economy and people's livelihood.The accuracy and stability of housing price prediction have important practical significance for the government,developers,and the general public.This paper uses the web crawler method to obtain the sample housing price data of the Anjuke real estate information service platform in Yinchuan from April 2015 to March 2021,and respectively uses Lasso model and Gradient Elevation Decision Tree(GBDT)model to predict the housing price in Yinchuan;Combining the advantages of the two prediction methods,a Lasso GBDT combined regression prediction model based on Stacking ensemble learning is constructed.By comparing the prediction results of examples,the combined prediction models have a prediction accuracy of over 0.98,which can effectively avoid the impact of pathological data on the fitting degree,and have higher accuracy and stability than single prediction models.