박민영, 강정은 (2024). 가로 단위 보행만족도 예측 모델 구축 및 부산시 적용: 의 미론적 분할 알고리즘과 머신러닝을 활용하여. 2024 대한국토도시계획학회 춘계학술대회. (2024. 5. 9. - 2024. 5. 10.) 

<Abstract>

 This study aimed to predict pedestrian satisfaction in Busan City using a model trained on survey data from Seoul City. To do so, we collected Google Street View images based on Seoul's pedestrian satisfaction survey locations and used semantic segmentation algorithms to identify physical environmental factors. Pixel ratios were calculated and utilized as independent variables. Additionally, we included slope, which is not easily discernible from images alone, as an independent variable. Before model training, we addressed class imbalance by conducting oversampling and trained 14 different models. Through validation on test data, we selected the model with the highest performance, the Extra Trees Classifier model, as the final model and applied it to Busan City. The results showed a higher proportion of 'dissatisfaction' with pedestrian paths in central Busan compared to the eastern and western regions, indicating a need for improvement in pedestrian environments, especially in central Busan, despite its high population density. This highlights the importance of improving pedestrian environments in central Busan. This study provides valuable insights into pedestrian satisfaction status at the street level in Busan City and attempts to predict pedestrian satisfaction using landscape images. 

 키워드 : 머신러닝, 의미론적 분할 알고리즘, 부산, 보행만족도, 구글 스트리트 뷰 

 사사 : 스마트시티