一种基于随机森林和Light GBM的房产估价模型  

A Real Estate Valuation Model Based on Random Forest and Light GBM

作　　者：冯梓豪 刘从军 FENG Zihao;LIU Congjun(School of Computer Science,Jiangsu University of Science and Technology,Zhenjiang 212000;Jiangsu KeDa Huifeng Technology Co.,Ltd.,Zhenjiang 212000)

机构地区：[1]江苏科技大学计算机学院,镇江212000 [2]江苏科大汇峰科技有限公司,镇江212000

出　　处：《计算机与数字工程》2024年第1期184-189,共6页Computer & Digital Engineering

摘　　要：针对商品房评估方法中存在数据源单一,考虑影响因素理想化,特征工程过分依赖主观经验等问题,结合随机森林和Light GBM模型,提出了一种RF_LightGBM模型用于房产价值评估。首先,通过随机森林对特征进行重要度排序,将影响房产价格因素较小的特征排除,使用网格搜索算法对模型进行优化,最后将该方法用于房产价值评估。在真实的房价数据集上进行的实验表明,相较于随即森林,XGBoost等传统模型,RF_LightGBM模型的评估精度提高了1.7%,且百分误差在0%~10%以内的评估结果占比88.38%。说明所用模型可以很好地应用于房产价值评估,得到的评估结果更加准确。In order to solve the problems of single data source,idealization of influencing factors and over reliance on subjec-tive experience in Feature Engineering in commercial housing evaluation,a new method based on random forest and light GBM mod-el is proposed.RF_Lightgbm model is used to evaluate real estate value.Firstly,the importance of features is sorted by random for-est,and the features that have little influence on the real estate price are excluded.The grid search algorithm is used to optimize the model.Finally,the method is applied to the real estate value evaluation.Experiments on real house price data sets show that,com-pared with traditional models such as random forest and XGboost,RF is better.The accuracy of lightgbm model is improved by 1.7%,and the percentage error within 0%~10%accounts for 88.38%.It shows that the model can be well applied to the real estate value evaluation,and the evaluation results are more accurate.

关 键 词：随机森林 房产估价 特征工程 Light GBM 

分 类 号：O141.4[理学—数学]