基于LSTM-StackingCXR模型的房价预测算法研究

Research on House Price Prediction Algorithm Based on LSTM-StackingCXR Model

房屋价格一直是社会最为关心的话题.预测房价走势并为购房者提供参考房价一直是房地产业和相关学术领域的研究热点.针对房价预测过程中存在数据集变量多、维度高的问题,本文通过计算多个房源特征与房源价格之间的皮尔森系数,去除冗余房源特征,有效地降低了房源特征数据集的维度.为了将信息损失降至最小,采用CatBoost处理房源特征中的类别变量.针对预测模型的过拟合、泛化能力差的问题,采用Stacking策略融合了CatBoost、XGBoost、随机森林算法,并且结合LSTM神经网络,最终提出了一种LSTM-StackingCXR模型.实验结果表明,LSTM-StackingCXR模型预测结果与现有多个模型预测结果相比,其预测精度指标有较大程度的提升.

Housing prices has always been the most concerning topic in society.Predicting the trend of housing prices and providing reference for homebuyers has always been a research hotspot in the real estate industry and related academic fields..In response to the problem of multiple variables and high dimensionality in the dataset of housing price prediction,this article calculates the Pearson coefficient between multiple housing features and housing prices,removes redundant housing features,and effectively reduces the dimensionality of the housing feature dataset.In the process of data preprocessing,a Catboost category variable processing method is used to minimize the information loss.In view of the problem of overfitting and poor generalization ability of the prediction model,a LSTM-StackingCXR model is established by combining stackingstrategy with Catboost,XGBoost and random forest model.The experimental results show that the prediction results of LSTM-StackingCXR model has significantly improved compared to the prediction results of multiple existing models.