http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
주거지역 가로환경 및 일상 걷기가 정신 건강에 미치는 영향 - 서울시 대상으로 -
구본유,백승주,윤희연,Koo, Bonyu,Baek, Seungjoo,Yoon, Heeyeun 한국조경학회 2024 韓國造景學會誌 Vol.52 No.1
This study aimed to investigate the impact of the quality of the street environment in residential areas on the mental health of urban residents, considering the frequency of street use. Using a zero-inflated negative binomial regression model, the study analyzed the influence of walking frequency and the street environment on depressive symptoms of urban residents. The research focused on Seoul, South Korea, in 2017, with depressive symptoms as the dependent variable and street environment variables, walking variables, and individual characteristics as independent variables. Additionally, the study explores the interaction effect of street greenery and walking frequency to analyze the synergistic impacts of walking in green spaces on mental health. The findings indicate that a higher ratio of street green areas is associated with fewer depressive symptoms. Increased walking frequency is linked to a reduction in depressive symptoms or a weaker manifestation of such symptoms. The interaction effect confirms that more frequent walking in green spaces is associated with weaker depressive symptoms. Lower ratios of visual complexity are correlated with reduced depressive symptoms. This study contributes to addressing urban residents' mental health issues at the community level by emphasizing the importance of the street green environment in residential areas.
진동 신호의 스펙트로그램을 통한 CNC 공구 마모도 예측
강재민,임현진,구본유,권선영 한국정보과학회 2023 정보과학회 컴퓨팅의 실제 논문지 Vol.29 No.8
The use of appropriate tools in the CNC process greatly affects the overall cost of the plant and the quality of the product. Various methods are proposed for precise timing of CNC tool replacement. As a part of it, this research processed vibration time series data as an image through audio feature extraction and searched for ways to use vibration data. For predicting CNC tool wear, the accuracy was only 84.68% when using the 1D-CNN model for time-series data, whereas the accuracy was 93.75% when using the 2D-CNN model for extracted vibration images. Furthermore, the performance was improved up to 94.61% when signals from different axes were used simultaneously, and an accuracy of 98.33% was obtained when images and time series data were used simultaneously. Vibration data can be used as a useful feature for determining tool wear, and it is expected that learning methods using this feature will contribute more to tool wear management in the future.
분자 그래프 특징 분석 및 수소와 원자 간 결합 정보 사용에 따른 성능 비교
노다솜,손연경,구본유,권선영 한국정보과학회 2024 정보과학회 컴퓨팅의 실제 논문지 Vol.30 No.1
Graphs are used to represent data emphasizing relationships between entities. These graph datasets exhibit diverse characteristics based on their application domains. Therefore, it is crucial to consider these attributes when developing artificial intelligence models. In this paper, our objective was to analyze the characteristics of molecular graph datasets commonly utilized for developing models that predict molecular properties, through a comparison with graph datasets from diverse fields. Furthermore, we aim to identify an effective feature composition for molecular property prediction through performance comparisons based on the utilization of hydrogen and interatomic bonding information in molecular graphs. As a result, molecular graphs exhibited distinctive characteristics compared to other fields. Utilizing hydrogen information as node attributes and incorporating interatomic bonding information as edge attributes improved the performance of the model.