• 한국경영정보학회:학술대회논문집
• /
• 2007.06a
• /
• pp.135-140
• /
• 2007

이탈 고객 예측은 데이터 마이닝에서 다루는 주요한 문제 중에 하나이다. 이탈 고객 예측은 일종의 분류(classification) 문제로 의사결정나무추론, 로지스틱 회귀분석, 인공신경망 등의 기법이 많이 활용되어왔다. 일반적으로 이탈 고객 예측을 위한 모델은 고객의 인구통계학적 정보와 계약이나 거래 정보를 입력변수로 하여 이탈 여부를 목표변수로 보는 형태로 분류 모델을 생성하게 된다. 본 연구에서는 고객과의 지속적인 접촉으로 발생되는 추가적인 사건 정보를 활용하여 연관성 규칙을 생성하고 이 결과를 기존의 방식으로 생성된 분류 모델과 결합하는 이탈 고객 예측 방법을 제시한다. 제시한 방법의 유용성을 확인하기 위해서 특정 국내 신용카드사의 실제 데이터를 활용하여 실험을 수행하였다. 실험 결과 제시된 방법이 기존의 전통적인 분류 모델에 비해서 향상된 성능을 보이는 것을 확인할 수 있었다. 제시된 예측 방법의 장점은 기존의 이탈 예측을 위한 입력 변수들 이외에 고객과 회사간의 접촉을 통해서 생성된 동적 정보들을 통합적으로 활용하여 예측 정확도를 높이고 실시간으로 이탈 확률을 갱신할 수 있다는 점이다.

• Journal of Korea Game Society
• /
• v.14 no.5
• /
• pp.35-42
• /
• 2014

In the market of First Person Shooter (FPS) games, Massively Multi-player Online FPS games (MMOFPS) like 'PlanetSide 2' have been popular recently. Dead reckoning has been widely used in order to mitigate the network traffic overload for the game server with hundreds or thousands of people. This paper proposes the efficient analytical method to calculate the tolerable threshold angle of moving direction, which is one of the most important factors for character status updating when dead reckoning is used in MMOFPS games. The experimental results with game testers shows that the proposed method minimizes the position error for character moving and provides natural direction updates of characters.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.31 no.12C
• /
• pp.1280-1287
• /
• 2006

We propose a modified stochastic gradient-based (MSGB) filter showing that the filter is the solution to an optimization problem. This paper analyzes the properties of the MSGB filter that corresponds to the nonlinear adaptive filter with additional update terms, parameterized by the variable scaling factor. The variably parameterized MSGB filter plays a role iii connecting the fixed parameterized MSGB filter and the null parameterized MSGB filter through variably scaling parameter. The stability regions and misadjustments are shown. A system identification is utilized to perform the computer simulation and demonstrate the improved performance feature of the MSGB filter.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2023.05a
• /
• pp.233-233
• /
• 2023

효율적인 수자원 관리에 필수적인 요소 중 하나는 유역 유출의 정확한 예측이다. 동일한 유역이라 할지라도 과거 기후조건에 대해 매개변수나 모형구조가 최적화된 수문모형은 현재나 미래 기후에 대해 최적이라 할수 없으며, 이에 따라 유역 유출 해석의 불확실성 또한 증가하고 있다. 수문자료동화는 모형의 입력 자료에 따른 불확실성을 줄이고 예측정확도를 향상 시킬 수 있는 방법으로, 수문모형의 상태량이나 매개변수를 업데이트하여 모형 초기 조건의 가능성 높은 추정치를 생성하는 기법이다. 본 연구에서는 국내 댐 상류 유역에 대해 집중형 수문모형과 순차자료동화 기법의 연계 패키지인 airGRdatassim 모형을 적용하여, 앙상블 칼만 필터와 파티클 필터 기법의 수문자료동화 특성을 비교 분석하고, 자료동화와 관련된 하이퍼-매개변수의 불확실성이 수문모의 성능에 미치는 영향을 분석하였다. 자료동화 적용 결과, 두 자료동화 기법 중 파티클 필터에 의한 모의성능이 높았으며 기상강제력 노이즈의 범위, 갱신 대상 상태량 설정, 앙상블 설정 등 수문자료동화의 설정과 관련된 하이퍼 매개변수의 불확실성은 두 기법별 뚜렷한 차이를 보였다. 또한, 본 연구에서는 일단위에서 시단위로 확장한 유량 예측 자료동화의 시험 모의결과 및 앙상블 수문동화기법의 도전과제에 대해서도 논의한다.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.28B no.10
• /
• pp.771-782
• /
• 1991

In industrial multivariable plants, it is ofte the case that the plant outputs are measured in similar components not simultaneously but serially. In this paper, the problem of estimating the state vector of the plant based on the data obtained from such a measuring scheme is considered, and a special type of observer(referred to as a $'$multiple serial-sampling$'$ type observer) which renews its internal states whenever a new group of data is obtained is proposed. It is proved that such an observer can be constructed for almost every sampling period if the palnt is observable as a continuous-time multivariable system, and that the poles of the closed-loop system using the multiple serial-sampling type observer consist of the poles of the observer and those of the state feedback system. The behaviors of the observer and the closed-loop system are studied by simulation. The results of simulation indicate that a multiple serial-ampling type observer can estimate the state of the plant more accurately than the ordinary type observers and improve the closed-loop performance, especially, in the existence of measuring noise.ng noise.

• Water for future
• /
• v.24 no.3
• /
• pp.83-94
• /
• 1991

In this paper a discrete extended Kalman filter for the tidal prediction has been developed. The filter is based on a set of difference equations derived from the one dimensional shallow water equations using the finite difference scheme proposed by Lax-Wendroff. The filter gives estimates of the water level and water velocity, together with the parameters in the model which essentially have a random character, e.g. bottom friction and wind stress. The estimates are propagated and updated by the filter when the physical circumstances change. The Kalman-filter is applied to field data gathered in the coastal area alon the West Sea and it is shown that the filter gives satisfactory results in forecasting the waterlevels during storm surge periods.

• Journal of the Korean Society for Precision Engineering
• /
• v.22 no.5 s.170
• /
• pp.45-54
• /
• 2005

The purpose of this research is to find a simple and accurate NURBS interpolator for CNC systems such as robot, CMM and CNC machine tools. This paper presents a new design of NURBS interpolator for CNC system. The proposed algorithm used the recursive characteristics of NURBS equation, the previous incremental value and chord length for the sake of a constant chord length. Simulation study was conducted to see the performance of the proposed interpolator with reference-word and reference-pulse method. Consequently, an accurate and simple NURBS interpolator was possible for modem CNC systems.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.2
• /
• pp.189-194
• /
• 2020

This paper analyzes the gradient descent method, which is the one most used for learning neural networks. Learning means updating a parameter so the loss function is at its minimum. The loss function quantifies the difference between actual and predicted values. The gradient descent method uses the slope of the loss function to update the parameter to minimize error, and is currently used in libraries that provide the best deep learning algorithms. However, these algorithms are provided in the form of a black box, making it difficult to identify the advantages and disadvantages of various gradient descent methods. This paper analyzes the characteristics of the stochastic gradient descent method, the momentum method, the AdaGrad method, and the Adadelta method, which are currently used gradient descent methods. The experimental data used a modified National Institute of Standards and Technology (MNIST) data set that is widely used to verify neural networks. The hidden layer consists of two layers: the first with 500 neurons, and the second with 300. The activation function of the output layer is the softmax function, and the rectified linear unit function is used for the remaining input and hidden layers. The loss function uses cross-entropy error.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.45 no.3
• /
• pp.11-19
• /
• 2008

Background subtraction, which separates moving objects in video sequences, is an essential technology for object recognition and tracking. However, background subtraction methods are often confused by shadow regions and this misclassification of shadow regions disturbs further processes to perceive the shapes or exact positions of moving objects. This paper proposes a method for shadow elimination which is based on shadow modeling by color information and Bayesian classification framework. Also, because of dynamic update of modeling parametres, the proposed method is able to correspond adaptively to illumination changes. Experimental results proved that the proposed method can eliminate shadow regions effectively even for circumstances with varying lighting condition.

• Journal of Korea Water Resources Association
• /
• v.44 no.10
• /
• pp.843-851
• /
• 2011

In hydrology, it is appropriate to use probabilistic method for forecasting mid/long term streamflow due to the uncertainty of input data. Through this study, it is expanded mid/long term forecasting system more effectively adding priory process function based on PDF-ratio method to the RRFS-ESP system for Guem River Basin. For implementing this purpose, weight is estimated using probabilistic weather forecasting information from KMA. Based on these results, ESP probability is updated per scenario. Through the estimated result per method, the average forecast score using ESP method is higher than that of naive forecasting and it confirmed that ESP method results in appropriate score for RRFS-ESP system. It is also shown that the score of ESP method applying revised inflow scenario using probabilistic weather forecasting is higher than that of ESP method. As a results, it will be improved the accuracy of forecasting using probabilistic weather forecasting.