서울시 범죄 토픽의 시공간 변화 분석 : 빅카인즈 뉴스 데이터와 KoBERTopic 모형을 활용하여

Abstract

본 연구는 서울시를 대상으로 장기간(2000-2023)에 해당하는시계열 데이터 구축을 통해 범죄 동향의 변화를 분석하고자 한다. 이를 위해 24년간 뉴스 데이터를 크롤링을 진행하였으며, 시· 공간적으로 범죄 동향의 변화를 파악하고자 KoBERTopic을 활용한 시계열 토픽모델링과 핫스팟 분석을 진행하였다. 시간의 흐름에 따른 도시 내 공간 변화를 확인하기 위해 네트워크 분석방법론을 활용하여 범죄 키워드 사이의 관계를 도출하고자 한다. 본연구의 결과는 범죄 유형과 동향을 키워드 도출을 통해 효율적으로 파악할 수 있고, 범죄 핫스팟 지역의 변화를 파악하여 이에 따른 안전한 도시를 위한 정책 형성에 유용하게 활용될 수 있다.

Crime negatively impacts daily activities and diminishes the quality of life for urban residents. Analyzing the spatial distribution and types of crime, along with formulating strategies to reduce crime, is essential for enhancing citizens' well-being. In Korea, crime location data is private, while crime rate data, categorized by type, is available at the city, county, and district levels. Recently accumulated digital news data provides details on crime locations, types, and targets. However, extracting relevant crime-related information from the extensive collection of news articles amassed over the past decades has posed challenges. This study examined the spatial distribution and changes in major crimes by analyzing news article data from 2000 to 2023, sourced from Big Kinds, utilizing the KoBERTopic text mining methodology and focusing on Seoul. Crime-related news was filtered using text mining technology, and location information tied to administrative districts was extracted from the articles to analyze crime concentration over time. Moreover, shifts in major crime topics were identified using the KoBERTopic methodology. The analysis revealed that the types of crimes and crime-concentrated areas in Seoul changed annually, influenced by significant factors. Furthermore, changes in crime topics over time were visually represented spatially. This study is notable for leveraging large-scale news data to analyze shifts in crime locations and concentrations, leading to policy recommendations for crime prevention.