논문 상세보기
pp.147-159
This study quantitatively investigates the structural impacts of weather and air pollution factors on bus mode choice probability in Seoul. Time-series panel datasets from 2022 to 2024 were constructed by integrating hourly bus ridership data with meteorological and air quality measurements. A random forest regression model was utilized to hierarchically evaluate variable importance and directionality, while monthly independent models and k-means clustering were applied to interpret seasonal dynamics and behavioral response patterns. The empirical results indicate that the discomfort index, relative humidity, particulate matter, and ozone consistently exhibited high variable importance throughout the year, serving as crucial factors in determining bus mode selection. In particular, extreme weather conditions during the hottest and coldest months worsened the perceived outdoor thermal stress, significantly reducing the bus mode share. Conversely, precipitation, carbon monoxide, and sulfur dioxide acted as auxiliary variables with relatively low importance across all months. Directionality analysis revealed that most environmental factors had a negative correlation with the bus mode share, reflecting commuters' aversion to outdoor exposure at bus stops. However, relative humidity and precipitation demonstrated positive relationships, indicating a mandatory constraint mechanism where travelers inevitably select buses over long-distance walking owing to dense bus stop accessibility under inclement weather. These findings imply that meteorological and air quality changes are structurally and seasonally embedded in bus transit demand, highlighting the potential for predictive, climate-responsive traffic operations. Utilizing key environmental indices, such as humidity and discomfort levels, to dynamically adjust bus-dispatch intervals could serve as an efficient public transportation management strategy in the era of climate change.
1. 서론
1.1. 연구의 배경 및 범위
1.2. 연구의 목적
2. 이론적 배경 및 선행연구 검토
2.1. 랜덤 포레스트의 원리
2.2. K-means 클러스터링 기반 유형화
2.3. 선행 연구 검토
3. 분석 방법론
3.1. 연구 수행 절차
3.2. 자료 수집 및 전처리
3.3. 변수 구성 및 다중공선성 점검
3.4. 변수 중요도 산정
3.5. 영향 방향성 분석
4. 분석 결과
4.1. 모형 성능 평가
4.2. 서울특별시 전체 변수 중요도 및 방향성
4.3. 월별 변수 중요도 변화
4.4. 군집 형성 및 계절 비교
5. 결론
5.1. 분석 결과 요약
5.2. 정책적 시사점
5.3. 연구의 한계 및 향후 과제
REFERENCES
