改进的XGBoost模型在短租房价格预测中的应用  

The Application of improved XGBoost in the Prediction of Short-Term Rental Housing Price

作　　者：郑列[1] 穆新宇 ZHENG Lie;MU Xinyu(School of Sciences,Hubei Univ.of Tech.,Wuhan 430068,China)

机构地区：[1]湖北工业大学理学院,湖北武汉430068

出　　处：《湖北工业大学学报》2021年第2期104-109,共6页Journal of Hubei University of Technology

基　　金：教育部人文社会科学研究规划基金项目(17YJA790098)。

摘　　要：对短租房价格原始数据集进行缺失值和异常值处理,针对短租房价格的影响因素构建包括23个特征的特征体系,使用OLS回归和分位数回归对这些因素的影响程度和影响方向进行分析,最后挑选具有较强显著性的18个特征构建XGBoost模型,用于预测房源价格。建模过程中采用网格搜索法调参。拟合优度这一指标在使用XGBoost模型进行价格预测时可以达到0.60,而线性回归模型仅为0.38。因此,使用XGBoost模型对短租房价格进行预测较优,将其与OLS回归和分位数回归相结合,既保留了传统统计模型的解释性,又提升了预测的精确度。This paper studies the prediction model of the housing price via the information of listings.Firstly,it processes the missing and abnormal values of the original data.Secondly,it constructs a reasonable feature system including 23 features for the influencing factors of short term rental housing price.Thirdly,it uses OLS regression and quantile regression to analyze the influence of these factors.Finally,it selects 18 outstanding features to construct the XGBoost model for predicting the price of listings.The model uses the grid search method to adjust parameters.The goodness of fit of XGBoost is 0.60,while that of linear regression is only 0.38.Therefore,the XGBoost combined with OLS regression and quantile regression,not only keeps the interpretation of the traditional model,but also improves the prediction accuracy.

关 键 词：在线短租房 OLS回归 分位数回归 XGBoost 网格搜索法 

分 类 号：F299.23[经济管理—国民经济]