Đánh giá các yếu tố ảnh hưởng tới quy hoạch sử dụng đất bằng phương pháp học máy

Abstract

Evaluating the factors that influence land use planning implementation is a crucial challenging task, particularly when detailed assessments of each planned location on the map are required. Machine learning, with its ability to process large, diverse datasets and detect complex nonlinear relationships, has proven effective across various fields. This study proposes a procedure using the Random Forest machine learning model to evaluate the factors affecting land use planning implementation. The dataset consists of 474 samples representing planned location for the 2010–2020 period in four districts of Son La Province, with 19 influencing variables. Random Forest (RF) outperformed Support Vector Machine (SVM) and Decision Tree (DT) on four of five performance metrics. Accordingly, we employed SHAP (SHapley Additive exPlanations) on the RF model to quantify variable contributions. The most significant factors affecting the success of land use planning implementation include: proximity to residential area, land use planing type, the planned land area, and elevation. This study presents a novel approach to land use planning assessment, enhancing objectivity, reliability, and ultimately improving the effectiveness of land use planning practices.