Wang Lei, Liang Yujie, Shang Gaizhen, Song Zhiyong, Gao Peng
NORENDAR International Ltd., Shijiazhuang 050011, Hebei, China.
Urban and Rural Construction College, Agricultural University of Hebei, Baoding 071066, Hebei, China.
Comput Intell Neurosci. 2022 Jun 10;2022:7936522. doi: 10.1155/2022/7936522. eCollection 2022.
This exploration aims to promote the development of urbanization in China and improve the utilization rate of urban resources. First, intensive theory and spatial economics are studied. Next, an input-output urban spatial evaluation system is established based on intensive theory and data envelopment analysis (DEA). Then, deep learning (DL) is adopted for optimization, and an urban space evaluation system based on DL is proposed. Finally, the reliability level of the urban space evaluation system is tested. The results show that the model's input and output index values are above 0.9, and the overall reliability level is higher than 0.9, indicating that the urban space evaluation system has a high reliability. The training results of the DL model show that the mean absolute error (MAE) of model prediction decreases gradually with the increase of training time and training times. When the training lasts for 5 min, each index' MAE is basically stable between 0.22 and 0.23, and the evaluation accuracy is obvious. The urban space evaluation system based on DL has higher evaluation accuracy, reaching 83.40%. Therefore, this exploration can provide research experience for promoting the effective utilization of urban resources and provide a reference for formulating an urbanization evaluation index system suitable for China's national conditions.
本探索旨在推动中国城市化发展，提高城市资源利用率。首先，研究集约理论和空间经济学。其次，基于集约理论和数据包络分析（DEA）建立投入产出城市空间评价体系。然后，采用深度学习（DL）进行优化，提出基于DL的城市空间评价体系。最后，测试城市空间评价体系的可靠性水平。结果表明，模型的输入和输出指标值均高于0.9，整体可靠性水平高于0.9，表明城市空间评价体系具有较高的可靠性。DL模型的训练结果表明，模型预测的平均绝对误差（MAE）随着训练时间和训练次数的增加而逐渐减小。当训练持续5分钟时，各指标的MAE基本稳定在0.22至0.23之间，评价精度明显。基于DL的城市空间评价体系具有较高的评价精度，达到83.40%。因此，本探索可为促进城市资源的有效利用提供研究经验，并为制定适合中国国情的城市化评价指标体系提供参考。
