| 融合时序数据和面板数据的LSTM-RBF城区面积预测模型 |
| |
| 引用本文： | 田子陶,陈晓勇,张翰超.融合时序数据和面板数据的LSTM-RBF城区面积预测模型[J].测绘与空间地理信息,2020(5):116-120. |
| |
| 作者姓名： | 田子陶 陈晓勇 张翰超 |
| |
| 作者单位： | 东华理工大学测绘工程学院;武汉大学遥感信息工程学院 |
| |
| 摘 要： | 城市扩展模拟预测是将城市时空演变规律应用于城市规划建设决策的重要基础,也是城市研究的热点。现有研究主要分为两类:一类是从城市时空演变角度出发采用异速增长等模型利用时序数据预测;另一类从城市扩张驱动力角度出发采用各类神经网络或多元回归模型结合面板数据进行预测。这两类算法均从单一方面对城市扩张做出解释,缺乏对城市时空演变规律和驱动因素双方面的综合考虑。本文提出了融合时序数据和面板数据的LSTM-RBF城区面积预测模型,该模型通过将长短期记忆(LSTM)网络和径向基函数(RBF)网络相结合,实现了时间序列城区面积和经济、人口等驱动力数据双方面支持下的城区面积联合预测,提高了城区面积预测的精度,为城区面积预测提供了一种新颖有效的方法,能够为城市时空演变研究服务于城市土地利用和规划制订提供技术支持。
|
| 关 键 词： | 城区面积 预测 时序数据 面板数据 LSTM-RBF |
A LSTM-RBF Urban Expansion Prediction Model Based on Time Series Data and Panel Data |
| |
| TIAN Zitao,CHEN Xiaoyong,ZHANG Hanchao.A LSTM-RBF Urban Expansion Prediction Model Based on Time Series Data and Panel Data[J].Geomatics & Spatial Information Technology,2020(5):116-120. |
| |
| Authors: | TIAN Zitao CHEN Xiaoyong ZHANG Hanchao |
| |
| Institution: | (Faculty of Geomatics,East China Institute of Technology,Nanchang 330013,China;School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430079,China) |
| |
| Abstract: | Urban expansion prediction is an important basis for applying urban spatial and temporal evolution law to urban planning and construction decision-making,and is also a hot topic of urban research.There are two kinds of methods for urban expansion prediction.One is to use time series data to predict by allometric growth model or others.Another is to use panel data to predict by various neural networks or multiple regression models from the perspective of urban expansion driving force.These algorithms only explain urban expansion from a single perspective and lack of comprehensive considerations of both urban temporal and spatial evolution and driving factors.Thus,we proposed LSTM-RBF model based on time series data and panel data.It combines long-term and short-term memory(LSTM)networks with radial basis function(RBF)networks to predict urban areas under the support of time series of urban areas and driving data such as economy and population.Compared with LSTM networks and RBF networks,the accuracy of LSTM-RBF model is higher.This novel and effective method for urban area prediction can provide technical support for urban spatial and temporal evolution research to serve urban land use and planning. |
| |
| Keywords: | urban area prediction time series data panel data LSTM-RBF |
| 本文献已被 CNKI 维普 等数据库收录！ |
