딥러닝을 활용한 서울시 주택 유형별 전세가격에 영향을 미치는 요인분석

Abstract

1. CONTENTS
 (1) RESEARCH OBJECTIVES This study used deep learning DNN model to empirically analyze the factors affecting Chonsei prices of each housing type in Seoul.
 (2) RESEARCH METHOD The output variables are set as the apartment Chonsei price index, the single-family home Chonsei price index, and the townhouse Chonsei price index, and the input variables are the apartment sales price index, single-family home sales price index, townhouse sales price index, number of employed people, and corporate bond yield. housing construction permit records. The spatial scope was set to Seoul, and the temporal scope was set from January 2007 to October 2023.
 (3) RESEARCH FINDINGS Looking at the LIME(Local Interpretable Model-agnostic Explanations) of each Chonsei prices input variable by housing type, when the apartment sale price index is -0.66 or less, the single-family house sale price index is in the range of -0.66 to -0.38, and the townhouse sale price index is -0.59 or less. it was found to have a negative impact on the Chonsei price index for each housing type. When the corporate bond yield was in the range of -0.89 to -0.23 for apartments, -0.89 to -0.23 for single-family homes, and -0.89 to -0.23 for townhouses, it was found to have a positive (+) impact on the Chonsei prices index for each housing type.
 2. RESULTS
 The implications are that although there are differences by housing type, the Chonsei price appears to be greatly influenced by the sale price and interest rates. Therefore, the government must continuously monitor the Chonsei price and sale price by housing type to ensure housing stability for the working class and middle class and appropriate liquidity management.