LSTM-GCN-Bee Model for Classifying Ecological Drivers of Urban Green Space Changes: A Study on Urban Ecosystem Dynamics

Abstract

Urban green space is an important part of the urban ecosystem, providing key ecological services such as climate regulation, biodiversity support, and leisure space for urban residents. However, the rapid expansion of urban areas poses a significant threat to the extent and quality of these green spaces. To address the complexity of understanding of the drivers of change in urban green space, this study proposes an LGB model that combines long and short-term memory network (LSTM), graph convolutional network (GCN), and bee swarm algorithm (BCA). The LGB model aims to systematically classify and identify key ecological factors influencing the dynamics of urban green space by capturing temporal variation and spatial relationships in urban ecosystems. We utilized four diverse datasets, including satellite image data provided by Google Earth Engine (time span: 2010-2020), European Urban Atlas dataset, SEDAC socioeconomic data, and The Atlas of Urban Expansion data set. Experimental results show that the LGB model showed high accuracy on all datasets, especially 97.88% on the The Atlas of Urban Expansion dataset. Moreover, the LGB model also outperforms existing methods on multiple metrics including recall, F1 score and AUC value, and the confidence intervals for all metrics show the reliability of the model results. Although the LGB model shows superior performance, we are also aware that the model may have some limitations in different urban areas and data quality. Nevertheless, this study provides an advanced analytical tool for urban ecological management, providing important insights for urban planners and policy makers to help promote sustainable urban development and promote the healthy development of urban ecosystems through better green space conservation and planning.