• 지능정보연구
• /
• 제15권3호
• /
• pp.1-15
• /
• 2009

데이터 불균형 문제는 분류 및 예측 문제에서 하나의 범주에 속하는 표본의 수가 다른 범주들에 속하는 표본 수에 비하여 현저하게 적을 경우 나타난다. 데이터 불균형이 심화됨에 따라 범주 사이의 분류 경계영역이 왜곡되고 결과적으로 분류자의 학습성과가 저하되는 문제가 발생한다. 본 연구에서는 데이터 불균형 문제를 해결하기 위하여 Geometric Mean-based Boosting (GM-Boost) 알고리즘을 제안하고자 한다. GM-Boost 알고리즘은 기하평균 개념에 기초하고 있어 다수 범주와 소수 범주를 동시에 고려한 학습이 가능하고 오분류된 표본에 집중하여 학습을 강화할 수 있는 장점이 있다. 기업부실 예측문제를 활용하여 GM-Boost 알고리즘의 성과를 검증한 결과 기존의Under-Sampling, Over-Sampling 및 AdaBoost 알고리즘에 비하여 우수한 분류 정확성을 보여주었고 데이터 불균형 정도에 관계없이 견고한 학습성과를 나타냈다.

• 전산구조공학
• /
• 제10권1호
• /
• pp.173-183
• /
• 1997

본 논문에서는 랜덤한 축대칭 기하학적 초기결함을 갖는 원통이 축방향 충격하중을 받는 경우의 반경방향 변위가 임계기준치를 최초로 통과하는 확률론적 충격좌굴 파괴시간을 해석할 수 있는 방법을 제시하였다. 랜덤한 기하학적 초기결함의 생성을 위해 초기결함의 평균함수 및 상관함수를 이용하여 확률장을 형성하는 방법을 사용하였다. 본 논문에서 제시된 방법은 실제적인 기하학적 초기결함이 갖는 불확실성을 취급하는데 적절하고 실용적이므로 이를 고려한 원통의 구조안전도해석에 이용할 수 있다.

• 한국산업보건학회지
• /
• 제25권4호
• /
• pp.493-505
• /
• 2015

Objectives: The purpose of this study was to evaluate the relevance of adjusting a urinary sample for urine hippuric correction value and its effects. Urinary biological monitoring data are typically adjusted to a constant creatinine and specific gravity concentration to correct for variable dilutions among spot samples. This study was conducted to evaluate the suitability of adjusting the urinary concentrations of urine creatinine and specific gravity(SG). Methods: We measured the concentrations of hippuric acid, in spot urine samples collected from control(119), case(120) individuals. The value of hippuric acid was adjusted by SG and urinary creatinine(HPLC & Jaffe). Results: The major results were as follows. The concentrations of urinary creatinine and SG for the control group were 1.84 g/L(SD 0.99) for arithmetic mean and 1.56 g/L(GSD 1.86) for geometric mean by HPLC method, 1.57 g/L (SD, 0.82) for arithmetic mean and 1.33 g/L(GSD 1.85) for geometric mean by Jaffe method, 1.028(SD 0.09) for arithmetic mean and 1.02(GSD 1.06) for geometric mean by refractometer. Hippuric acid levels were 0.40 g/L(SD 0.51) by arithmetic mean and 0.20 g/L(GSD 3.59). In that case the exposed group was 1.40 g/L(SD 0.58) for arithmetic mean and 1.28 g/L(GSD 1.55) for geometric mean by HPLC method, 1.27 g/L(SD 0.56) for arithmetic mean and 1.14 g/L(GSD 1.62) for geometric mean by Jaffe method, 1.045 L(SD 0.27) for arithmetic mean and 1.02(GSD 1.13) for geometric mean by refractometer(P<0.05). Hippuric acid levels were 0.67 g/L(SD 0.79) for arithmetic mean and 0.39 g/L(GSD 2.94)(p<0.05). The urine creatinine concentrations were affected by gender(p < 0.01) but SG levels were not affected by gender or age(p>0.05). After adjustment, urine hippuric acid was correlated with creatinine(HPLC & Jaffe)(r=0.723, P<0.05, r=0.708, P<0.05) and SG(r=0.936, P<0.05) and the control group shows significantly higher than the case group. In the case group for adjusted urine hippuric acid was correlated with creatinine(HPLC & Jaffe), (r=0.736, P<0.05), r=0.549, P<0.05), SG(r=0.549, P<0.05). After adjusting urine hippuric acid by urine creatinine(HPLC and Jaffe method) and specific gravity, significant associations were found between the control group and case group, respectively(r=0.832, P<0.05, r=0.845, P<0.05) and (r=0.841, P<0.05, r=0.849, P<0.05). Specific gravity adjustment appears to be more appropriate for variations in the urine creatinine method. Conclusion: we found that urinary creatinine concentrations were significantly affected by gender, and other factors and that care should therefore be exercised when correcting urinary metabolites according to the urinary creatinine concentration in spot urine. It is determined that additional study is needed for biological monitoring.

• 한국자동차공학회논문집
• /
• 제5권5호
• /
• pp.197-212
• /
• 1997

The performance of a torque converter can be expressed by the performance parameters such as flow radius and flow angle, on the mean flow path. The geometric analysis of the torque converter is required to determine these parameters for the modeling of the torque converter. In general, the blade shape is depicted by three dimensional data at the mid-surface of blade or those of the pressure and suction side. To generate three dimensional model of the blade using the data mentioned above, a consistent data format and a shape generation algorithm are required. This paper presents a useful consistent data format of the blades and an algorithm for the geometrical shape generation. By the geometric analysis program to which the shape generation algorithm is embedded, the variation of blade angles in rotating element analyzed. Then finally, the analyzed results of geometric profile of a blade are compared with those of the blade design principle, so called forced vortex theorem.

• International Journal of Reliability and Applications
• /
• 제12권1호
• /
• pp.49-60
• /
• 2011

The geometric process (GP) has been widely used for modeling failure and repair time sequences of repairable systems. The GP is mathematically tractable but restrictive in reliability applications since it actually assumes that the mean function of inter-failure times sequence asymptotically decreases to zero; and the mean function of successive repair times sequence asymptotically increases to infinity. This is generally unrealistic from an engineering perspective. This paper presents three extended GP models for modeling reliability deterioration and improvement (or growth) process. The extensions maintain the advantage of mathematical tractability of GP model. Their usefulness and appropriateness are illustrated with three real-world examples.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2004년도 학술대회 논문집 정보 및 제어부문
• /
• pp.218-220
• /
• 2004

Orthogonal frequency division multiplexing(OFDM) is one of the most promising technique for next generation wireless broadband communication systems. In this paper, we propose a new bit allocation algorithm in multiuser OFDM. The proposed algorithm is derived from the geometric progression of the additional transmit power of subcarriers and the arithmetic-geometric means inequality. The simulation shows that this algorithm has similar performance to the conventional adaptive bit allocation algorithm and lower computational complexity than the existing algorithms.

• Kyungpook Mathematical Journal
• /
• 제47권4호
• /
• pp.595-600
• /
• 2007

In this paper we consider weighted arithmetic and geometric means of several positive definite operators proposed by Sagae and Tanabe and we establish a reverse inequality of the arithmetic and geometric means via Specht ratio and the Thompson metric on the convex cone of positive definite operators.

• 한국동물위생학회지
• /
• 제37권3호
• /
• pp.185-190
• /
• 2014

We investigated the serological prevalence of avian metapneumovirus (AMPV), avian reovirus (ARV), Mycoplasma gallisepticum (MG) and Mycoplasma synoviae (MS) in 760 broiler breeder (38 flocks), in the Jeonbuk province in 2013. This study was conducted to evaluate the immune and infection status of the broiler breeder flocks against AMPV, ARV, MG, MS by an enzyme-linked immunosorbent assay (ELISA). Serological test for AMPV were positive 37 (97.3%) flocks and 712 (93.6%) broiler breeder and geometric mean antibody titers were $16,350{\pm}10,195$, ARV were high positive rate 100% (38/38) flocks and 97.8% (743/760). The seropositive flocks against MG were 71.1% (27/38) and the geometric mean antibody titers were $2,474{\pm}2,045$, whereas the rates of positive flocks against MS were 50.0% (19/38) and the geometric mean antibody titers were $1,469{\pm}1,230$.

• East Asian mathematical journal
• /
• 제38권3호
• /
• pp.331-338
• /
• 2022

In this paper, we prove that the first nonzero eigenvalues λ₁ of the Laplacian and the p-Laplacian are decreasing along the inverse mean curvature flow with forced term in Euclidean space.

• 한국도로학회논문집
• /
• 제6권3호
• /
• pp.55-64
• /
• 2004

최대건조단위중량은 노상토 재료 특성으로서 매우 중요한 인자이다. 최대건조단위중량을 예측하는 기존의 모델들은 많은 변수를 포함하고 있어 다소 복잡해 보인다 본 논문에서는 사질토의 최대건조단위중량을 예측할 수 있는 간편한 식을 제안하였다. 이를 위해 36개 시료를 체분석하여 입도분포를 구하고, 다짐시험 한 후 그 결과를 회귀 분석하였다. 제안식은 변수로 노상토의 기하평균과 기하표준편차 또는 입도분포계수를 포함한다. 제안식의 검증을 위해 전국16개 지역의 채취 시료에 대한 최대건조단위중량의 실측치와 예측치를 비교한 결과 잘 맞는 것으로 밝혀졌다.