城市土地价格时空预测Stacking-GWR模型

Stacking-GWR Model for Spatio-temporal Prediction of Urban Land Prices

城市土地价格影响国土空间规划决策、现代城市治理和土地市场调控,预测城市土地价格具有重要意义,但不同用途的土地价格变化趋势差异显著且具有空间异质性,很难用单个模型进行预测。该文提出一种城市土地价格时空预测Stacking-GWR模型,以常州市主城区为研究区,根据土地价格变化趋势分为工业用地和非工业用地两组,利用Stacking-GWR模型进行土地价格预测,并与单独使用Stacking、地理加权回归(GWR)、时空地理加权回归(GTWR)模型的预测结果进行对比分析。结果表明:①Stacking-GWR模型融合了地价数据中的特征、空间和时间信息,能提高预测精度;②根据土地价格变化趋势进行分组后,模型预测精度优于不分组时的预测精度;③工业用地和非工业用地土地价格的全局和邻域影响因子差异显著。

The pricing of urban land significantly affects territorial and spatial planning,contemporary urban governance,and land market regulation,accurate forecasting of urban land prices is critically important.The price trends for different land uses exhibit notable differences and spatial heterogeneity,making it difficult to predict using a single model.Therefore,this paper establishes a new spatiotemporal prediction model for urban land prices,the Stacking-GWR model.The study focuses on the main urban area of Changzhou City,dividing land into industrial and non-industrial groups based on price trends.The Stacking-GWR model is used for land price prediction and compared with the results from using Stacking,geographically weighted regression(GWR),and geographically and temporally weighted regression(GTWR)models individually.The results are shown as follows.①The Stacking-GWR model,which integrates feature information,spatial information,and temporal information from land price data,improves prediction accuracy.②Prediction accuracy is higher for grouped predictions based on land price trends compared to non-grouped predictions.③There are significant differences in global and local influencing factors for land prices between industrial and non-industrial land.