基于随机森林与深度神经网络的房地产价值预测模型比较研究

打开文本图片集

中图分类号：TP391.9 文献标识码：A文章编号：1006-8228（2026）01-47-06

Abstract：Thisstudyadresesthelimitationsoftraditionalrealestateapraisalmethods，suchasstrongsubjectivityandlow eficiency，byproposingandimplementingamachinelearing-basedhigh-precisionpropertyvaluepredictionmodel.Theresearch beginswithsystematicdatapreproessngandfeaturescalingtoconstructastandardizedinputfeaturematrix.Onthisbasis，a comparativeframeworkisadoptedtoconstructthedeepneuralnetworkandrandomforestmodelsinparalel，whicharetrained andtestedunderaunifiedexperimentalenvironment.Experimentalresultsshowthattherandomforestmodelsignificantly outperformsthedeepneuralnetworkinpredictionperformance，achievingameanabsoluteerorof82，511.47yuanandamean absolute percentage error of 16.92% ，whiledemonstrating greater stabilityinpredicting lower-priced properties.

keywords：Real Estate Value Prediction；Random Forest;DeepNeural Network;Machine Learning

0引言

随着数字化和智能化的快速发展，房地产市场的复杂性愈加显著，传统的评估方法已难以满足当前房地产评估领域对准确性和效率的需求。（剩余6635字）