연결허가에 따른 공시지가 예측을 위한 심층신경망 기반 모델 개발

Abstract

This study aims to analyze the impact of road connection permits on land price and to develop a deep learning-based predictive model for land price fluctuations. A dataset of 1,225 cases from 2020 to 2022 was compiled by linking administrative records of road connection permits with official land price data for two years before and after each permit. Using this dataset, a deep neural network model based on a multilayer perceptron (MLP) was designed, and various activation functions (ReLU, LeakyReLU, PReLU, ELU, Tanh) were compared to identify the optimal configuration. The model was trained and evaluated using performance metrics such as MAE, RMSE, and MAPE. Experimental results demonstrated that ELU and PReLU achieved the most stable and accurate predictions, with average errors around 3%. The proposed approach highlights the feasibility of using deep learning to quantify the economic effects of road connec- tion permits and provides a data-driven foundation for improving fairness in fee calculation and supporting evidence-based policy decisions.