Machine Learning based land use classification and identification of driving factors in Google Earth Engine

Abstract

This thesis utilizes Google Earth Engine and machine learning to classify land use and analyze driving factors in Dalian, China, from 2000 to 2020. It employs a Random Forest algorithm on Landsat imagery to generate land use maps for four categories: agricultural land, vegetation, water, and built-up areas. Using Principal Component Analysis and Multiple Linear Regression, it identifies the primary drivers of land use change. The research demonstrates the effectiveness of cloud computing platforms for large-scale, long-term land use monitoring and provides a scientific basis for regional land management.