땅밀림 위험지 평가를 위한 기계학습 분류모델 비교

이제만 ( Jeman Lee ),서정일 ( Jung Il Seo ),이진호 ( Jin-ho Lee ),임상준 ( Sangjun Im ) 한국산림과학회 2021 한국산림과학회지 Vol.110 No.4

지진 발생과 집중호우에 의해 땅밀림형 산사태 유형으로 분류되는 땅밀림 현상이 전국적으로 광범위하게 나타나고 있다. 산림청은 땅밀림으로 인한 인명 및 재산 피해를 예방하기 위해 땅밀림 우려지 현장조사 판정표를 통해 땅밀림 발생 위험지를 사전에 파악하고 있다. 한편 최근에는 컴퓨터 기술의 발달로 인공지능의 한 분야인 기계학습 분류기법을 이용하여 산지재해 취약성을 평가하거나 자연재해를 예측하고 있다. 따라서 이 연구에서는 기계학습 분류기법인 k-Nearest Neighbor(k-NN), Naive Bayes(NB), Random Forest(RF), 그리고 Support Vector Machine(SVM) 분류모델을 이용하여 땅밀림 발생 위험등급을 분류하였다. 한국치산기술협회의 2018∼2020년 조사 자료 4,618개 중에서 땅밀림 현상의 발생 여부를 고려하여 발생지 총 146개소, 그리고 미발생지 146개소를 임의추출하여 292개 자료를 선정하였으며, 이 중 70%에 해당하는 204개소 자료를 훈련자료로 하여 모델을 구축하였다. 전체 자료의 30%에 해당하는 88개 검증자료에 대해 모델을 평가한 결과, k-NN은 0.727, NB는 0.750, RF는 0.807, 그리고 SVM은 0.750의 분류정확도를 보였다. 또한, Kappa 상관계수는 각각 0.534, 0.580, 0.673 및 0.585, 그리고 AUC는 각각 0.872, 0.912, 0.943 및 0.834로 계산되었다. 따라서 땅밀림 위험지역 판정을 위한 기계학습 분류모델은 RF, NB, SVM, 그리고 k-NN 순으로 높은 성능을 보였다. 기계학습 분류모델은 향후 산지토사재해의 예방 및 대응을 위한 기초자료로 활용 가능하며, 땅밀림 재해 관리 및 피해 경감에 위한 정책 개발에 필요한 정보를 제공할 것이다. The soil creep, primarily caused by earthquakes and torrential rainfall events, has widely occurred across the country. The Korea Forest Service attempted to quantify the soil creep susceptible areas using a discriminant value table to prevent or mitigate casualties and/or property damages in advance. With the advent of advanced computer technologies, machine learning-based classification models have been employed for managing mountainous disasters, such as landslides and debris flows. This study aims to quantify the soil creep susceptibility using several classifiers, namely the k-Nearest Neighbor (k-NN), Naive Bayes (NB), Random Forest (RF), and Support Vector Machine (SVM) models. To develop the classification models, we downscaled 292 data from 4,618 field survey data. About 70% of the selected data were used for training, with the remaining 30% used for model testing. The developed models have the classification accuracy of 0.727 for k-NN, 0.750 for NB, 0.807 for RF, and 0.750 for SVM against test datasets representing 30% of the total data. Furthermore, we estimated Cohen's Kappa index as 0.534, 0.580, 0.673, and 0.585, with AUC values of 0.872, 0.912, 0.943, and 0.834, respectively. The machine learning-based classifications for soil creep susceptibility were RF, NB, SVM, and k-NN in that order. Our findings indicate that the machine learning classifiers can provide valuable information in establishing and implementing natural disaster management plans in mountainous areas.

히스토그램 블록 기반 유사 영상 맵 생성 및 영상 합성 알고리즘

유재성(Jaeseong Yu),이은별(Eunbyeol Lee),김하린(Harin Kim),이재만(Jeman Lee),이의상(Euisang Lee),김규헌(Kyuheon Kim) 한국방송·미디어공학회 2016 한국방송공학회 학술발표대회 논문집 Vol.2016 No.6

본 논문에서는 다수의 영상을 빠르고 오류 없이 정합하기 위하여 정합과정의 전 처리로써 유사도 맵 생성 알고리즘을 제안한다. 본 논문에서는 블록화한 히스토그램을 통하여 영상간의 관계를 판별하게 된다. 두 영상의 블록 히스토그램을 비교하여 영상 간의 유사성과 위치관계를 8 방향으로 판별하고 이를 이용하여 유사도 맵에 영상들을 정렬하게 된다. 유사도 맵의 생성으로 정합 알고리즘을 적용해야 하는 경우의 수가 줄어들어 복잡도는 낮아지게 되어 이후 정합과정에서 속도의 이득을 얻을 수 있다. 또한 정합 방법으로 변형이 적은 영상을 정합하는데 탁월한 성능과 속도를 보이는 히스토그램을 이용한 방법을 제안한다. 제안 알고리즘을 이용하여 실험한 결과 기존의 다중 영상 스티칭 알고리즘에 비하여 매우 빠른 속도를 확인 할 수 있고 결과 영상 또한 오류가 적은 것을 확인 할 수 있다.

다표본 처리를 위한 새로운 AHP 활용 방안에 관한 연구

부제만(Boo, Jeman),이태원(Lee, Taewon) 글로벌경영학회 2015 글로벌경영학회 학술대회 발표논문집 Vol.2015 No.1

현대 사회에서 우리는 수많은 의사결정 상황에 봉착하게 된다. 그리고 우리가 선택하는 선택지로 인해 개인의 미래부터 나아가 조직의 미래가 결정된다. AHP(Analytic Hierarchy Process)는 이러한 선택의 순간 최선의 선택을 할 수 있도록 만드는 도구로서 유용하다. 하지만 개별적인 요인분석에 비해 다표본 분석에는 적합하지 않다. 그 이유로는 첫째, 표본 하나하나를 쌍대행렬로 변환하는 작업이 필요하고 둘째, 각각의 일관성을 확인할 수 있는 지표인 C.I(Consistency Index) 값의 분류가 쉽지 않다는 점이다. AHP 관련 연구들은 대부분 소규모의 수작업을 기본으로 하는 연구에 집중되어 있었으며, 답변들 역시 전문가를 대상으로 하는 경우가 많았다. 이에 본 연구는 이러한 점에서 다수의 일반인을 상대로 한 대량의 AHP 분석 데이터를 보다 빠르고 편리하게 결과를 낼 수 있을지를 시도하였으며, 이후 직접 계산 툴을 제작하여 목표한 연구 결과를 이끌 수 있었다. 본 연구는 기존의 어려운 방법으로 데이터를 넣어서 분석하던 어려움을 극복하고 빠른 시간에 많은 데이터를 처리할 수 있는 방법을 제시함으로써 연구자들의 많은 연구에 도움을 줄 수 있는것이 논문의 기여점으로 생각한다. In modern society, we are faced with the situation of a number of decisions . And because we choose the option from an individual s future is determined Moreover, the organization s future . AHP (Analytic Hierarchy Process) is the best choice at the moment this choice so that you can be useful as a tool . However, compared to the factor analysis of individual samples are not suitable for analysis . The first reason is that , one by one sample is necessary to convert the pairwise matrix , and second, you can see the consistency of each indicator CI (Consistency Index) is the value that is not easy classification . AHP-related research are the basis of most of the small hand had focused on research, were often targeted at professionals. In this regard, this study is helpful to the analyzing large number of AHP data more quickly and easily, and this method can making a direct calculation could lead to a study.

A User-Side Energy-Saving Video Streaming Scheme for LTE Devices

Song Gun Lee,Jeman Park,Hyogon Kim IEEE 2015 IEEE communications letters Vol.19 No.6

<P>Today, most streaming services rely on HyperText Transfer Protocol (HTTP) to deliver the contents to users, and mobile users are increasingly choosing Long Term Evolution (LTE) as the wireless access mode. In this letter, we propose a user-side scheme that takes LTE modem operation into account in video streaming over HTTP, leading to as much battery saving gain as 40% without Discontinuous Reception (DRX) and potentially 70% if DRX is deployed in future.</P>

Post Fire Effects on Structural and Hydrological Properties of Forest Soils

( Qiwen Li ),( Sangjun Im ),( Taehyun Kim ),( Jeman Lee ) 대한지질공학회 2019 대한지질공학회 학술발표회논문집 Vol.2019 No.2

A fire is one of the most dangerous events that lead to irreversible consequences on forest environment. All organic matters over and near the ground surface can burn in a fire. Combustion of forest fuels causes significant changes in soil structure and property. Fire-induced repellent layer plays an important role in controlling runoff and erosion processes after a fire. Therefore, an intensive survey on soil structure and hydrologic charactristics is needed to quantitatively examine the effects of fire on water movement and soil erosion. In this study, spatial and temporal patterns of soil properties were observed on forestlands in Gangwon province, affected by the 2019.04.04 fire. Soil texture, degrees of ash layer and soil water repellency were measured according to fire severity and fire type. Water Drop Penetration Time (WDPT) was implemented to measure the extent of soil water repellency for burnt soils. Through the field investigation, the results showed that soil properties changed with high heat, but the influences reached only a few centimeter near the soil surface. The extent and pattern of soil hydrophobicity have close relations to fuel types and fire intensities. Soil water repellency decreased with time, mainly due to raindrop impacts and soil sheet erosion. Acknowledgment This study was carried out with the support of ‘R&D Program for Forest Science Technology (Project No. 2017061B10-1919-AB01)’ provided by Korea Forest Service (Korea Forestry Promotion Institute)

Corrigendum to "PartitionTuner: An operator scheduler for deep-learning compilers supporting multiple heterogeneous processing units"

Misun Yu,Yongin Kwon,Jemin Lee,Jeman Park,Junmo Park,Taeho Kim Electronics and Telecommunications Research Instit 2023 ETRI Journal Vol.45 No.4

중소기업의 공급사슬망 통합수준이 SCM 성과 및 경영성과에 미치는 영향에 관한 연구

성호경(Ho-Kyung Sung),이민호(Minho Lee),부제만(Jeman Boo) 한국산업경영시스템학회 2020 한국산업경영시스템학회지 Vol.43 No.3

The purpose of this study is to examine the relationship between internal corporate, supplier, and customer integrations for domestic SMEs on non-financial and financial performance through SCM performance such as flexibility and reduction of uncertainties. To this end, data was collected on 286 SMEs in Korea, and the structural relationships between SCM integration level, SCM performances, and management performance were analyzed. As a result of the analysis, first, it was found that the SCM integration level had a significant positive effect on the flexibility and reduction of uncertainties, which are SCM performances. Second, the flexibility and reduction of uncertainties showed significantly positive effects on the non-financial performance of the companies, but did not directly affect the financial performance positively. Third, the non-financial performance was found to have a positive effect on the financial performance. In addition, the SCM integration level did not have a direct effect on the financial and non-financial performance, but it was found that it affected management performance by mediating the flexibility and reduction of uncertainties, which are SCM performances. That is, although the SCM integration level did not directly affect financial and non-financial performance, it was confirmed that it affects management performance by mediating SCM performances, flexibility and uncertainty reduction. In other words, it was confirmed that the SCM integration level directly or indirectly affects SCM performances and overall management performance. These results imply the necessity to focus on competency in the supply chain management area according to the SCM performance expected by SMEs, and the step by step approaches to the expected effects. In a situation where prior SCM related studies have not been able to present SCM performances and management performance of SMEs that are relatively lacking in their capital and SCM construction capabilities, the findings of this study could suggest the importance of SCM integration from the perspective of SMEs. In addition, from the viewpoint of SMEs, this study suggested that a sequential approach for performance measurement is required (SCM performance → management performance) in relation to the performance factors to be established through SCM.

Impact of Histologic Variants of Bladder Cancer on Oncology Outcome After Radical Cystectomy

Jae Hyeon Han,Se Young Choi,Sangjun Yoo,Seung-Hee Baek,Jeman Ryu,Yoon Soo Kyung,Wook Nam,Won Chul Lee,Dalsan You,In Gab Jeong,Bumsik Hong,Hanjong Ahn,Choung-Soo Kim 대한비뇨기종양학회 2017 대한비뇨기종양학회지 Vol.15 No.3

Purpose: To evaluate the oncological outcome of histologic variants in bladder cancer patients who underwent radical cystectomy. Materials and Methods: We identified 393 bladder cancer patients who underwent radical cystectomy at single center between January 2007 and August 2014. Patients were divided into 4 groups according to histologic types: pure urothelial cell carcinoma (UC) and squamous, micropapillary, and other variants. Kaplan-Meier analysis was performed to assess recurrence-free (RFS) and overall survivals (OS). The patients were divided into those with pathologic stage and nodal status. Results: Among 393 bladder cancer patients, squamous, micropapillary histologic variants were observed in 38 (9.7%), 26 (6.6%), respectively, whereas 39 had other variant types. Stage T3 cancer occurred in more patients with histologic variant compared with those with pure UC. Pathologic positive nodal status was also frequently found in the histologic variant groups. Subgroup analysis according to T stage and nodal status showed no significant difference in RFS and OS. On multivariate analysis, pathologic T stage (stage T2: hazard ratio [HR], 2.75; 95% confidence interval [CI], 1.34–5.63; p=0.005; stage ≥T3: HR, 3.20; 95% CI, 1.62–6.30; p=0.001) and nodal status (HR, 1.85; 95% CI, 1.05–2.56; p=0.028) were prognostic factors for RFS. Conclusions: In patients who underwent radical cystectomy, histologic variants were detected more often at advanced pathologic stage. Although histologic variants have been identified in the radical cystectomy specimen, treatment should be performed according to the pathologic stage.