Batch evaluation of collective owned commercialised construction land using machine learning.

Abstract

The market entry of collective owned commercialised construction land (CCCL) is a crucial element of China's ongoing rural land system reform. However, traditional appraisal methods struggle with efficiency and accuracy in the context of batch appraisals for CCCL market-entry prices. This study addresses this challenge by leveraging machine learning techniques to develop a batch appraisal model that enhances both efficiency and precision. Focusing on Beiliu City, a representative reform pilot area, we implemented three models-Random Forest (RF), Back Propagation Neural Network (BPNN), and Support Vector Machine (SVM)-to develop a tailored indicator system for price prediction. The results demonstrate that the RF model exhibits superior performance, achieving a mean absolute error of 17.50 yuan and a prediction accuracy of 94.77%, compared with 91.21% for BPNN and 91.94% for SVM. Moreover, this research reveals that CCCL prices display unique characteristics distinct from those of other land types, with significant influences from factors such as township economic levels and the specific approaches used for market entry. These findings validate the effective application of machine learning models in this context, offer a scientific foundation for standardising the land market, and serve as a guide for relevant policy formulation.