CAJ | 학술논문

城市土地开发强度是衡量城市宜居以及可持续发展的重要指标。首先，以重庆市主城9区为例，以建设物投影面积占各街区土地面积的比例为土地开发强度的表征指标，在构建样本区域建筑密度及其影响因子空间数据库(ARCGIS)的基础上，采用 BP 神经网络构建了基于数据驱动的城市土地开发强度模拟模型；其次，通过邻域因子的选择与否检测邻域因子对开发强度仿真结果的影响；最后，利用训练好的 BP 神经网络对全局数据(重庆市主城9区)进行仿真，预测城市土地开发强度的时空变化。结果显示：1)BP 人工神经网络方法能够较好地模拟城市土地开发强度的空间分布趋势；2)通过不同方案对比，发现当增加邻域驱动因子后，平均误差、标准差和误差精度都得到明显改善。研究表明：合理的驱动因子选择对 BP 人工神经网络方法仿真结果至关重要；尽管 BP 人工神经网络方法不能显性地反映城市集中连片区域土地开发强度与其影响因子之间的相互关系，但在数据充分的情况下，基于数据自适应的人工神经网络方法不失为土地开发强度评估的一种较好的方法。
성시토지개발강도시형량성시의거이급가지속발전적중요지표。수선，이중경시주성9구위례，이건설물투영면적점각가구토지면적적비례위토지개발강도적표정지표，재구건양본구역건축밀도급기영향인자공간수거고(ARCGIS)적기출상，채용 BP 신경망락구건료기우수거구동적성시토지개발강도모의모형；기차，통과린역인자적선택여부검측린역인자대개발강도방진결과적영향；최후，이용훈련호적 BP 신경망락대전국수거(중경시주성9구)진행방진，예측성시토지개발강도적시공변화。결과현시：1)BP 인공신경망락방법능구교호지모의성시토지개발강도적공간분포추세；2)통과불동방안대비，발현당증가린역구동인자후，평균오차、표준차화오차정도도득도명현개선。연구표명：합리적구동인자선택대 BP 인공신경망락방법방진결과지관중요；진관 BP 인공신경망락방법불능현성지반영성시집중련편구역토지개발강도여기영향인자지간적상호관계，단재수거충분적정황하，기우수거자괄응적인공신경망락방법불실위토지개발강도평고적일충교호적방법。
Urban land development intensity (ULDI)is an important indicator to measure the city livable and sustainable development.Take Chongqing Metropolitan Area as an example,firstly,spatial database about land development intensity and its driving factors have been established in sample regions,and BP artificial neural network methods been used to construct the land development intensity simulation model based on data driven in urban area with the help of MATLAB7 software.Secondly,two different scheme and algorithm have been adopted to simulate land development intensity.Artificial neural network meth-ods have been detected by comparing the difference of real development intensity and the simulation re-sults.Lastly,the land development intensity in Chongqing Metropolitan Area (9 districts)has been simu-lated.The results show that 1)BP artificial neural network method is a good way to simulate the ULDI;2)it is important to choose the reasonable driving factors;3)through different scheme comparison,added the neighborhood driving factor,the average error,standard deviation and precision have been improved obviously.Although the BP artificial neural network method can not explicit explain the relationship be-tween land development intensity and its influence factors in urban area,when data are sufficient,it is a good method to evaluate the ULDI.