• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2001.11a
• /
• pp.409-412
• /
• 2001

Identification of systems operating in closed loop has long been of prime interest in industrial applications. The fundamental problem with closed-loop data is the correlation between the unmeasurable noise and the input. This is the reason why several methods that work in open loop fail when applied to closed-loop data. The prediction error based approaches to the closed-loop system are divided to direct method and indirect method. Both of direct and indirect methods are known to be applied to the closed-loop data without critical modification. But the direct method induces the bias error in the experimental frequency response function and this bias error may deteriorates the parameter estimation performance

• Transactions on Control, Automation and Systems Engineering
• /
• v.4 no.4
• /
• pp.305-310
• /
• 2002

The measured rate of the tracking sensor becomes biased under some operational situation. For a highly maneuverable aircraft in 3D space, the target dynamics changes from time to time, and the Kalman filter using position measurement only can not be used effectively to reject the rate measurement bias error. To cope with this problem, we present a new algorithm which incorporate FIR-type filter and FIR-type fixed-lag smoother, and demonstrate that it has the optimal performance in terms of both estimation accuracy and response time through an application example to the anti-aircraft gun fire control system(AAGFCS).

• Journal of Magnetics
• /
• v.8 no.4
• /
• pp.160-163
• /
• 2003

In magnetic flux measurement, output voltage drift of electronic integrator is an essential problem. In this work, we have developed a new kind of Miller type integrator using a sample and hold amplifier. Input bias current was measured and this value was hold in the sample and hold amplifier, after that input bias current of Miller integrator was compensated automatically using the value which holds in the sample and hold amplifier. Developed flux-meter shows the drift of flux-meter are smaller than 10$^{-5}$ Wb/min in full scale of 10$^{-2}$, and we could also measure multi-channel magnetic flux simultaneously.

• The Korean Journal of Applied Statistics
• /
• v.29 no.2
• /
• pp.389-398
• /
• 2016

In this paper, we examine the financial performance of credit guarantee programs. We compared financial performance of guaranteed firms of KODIT and non-guaranteed firms. The of covariate adjusted propensity score method is used because a selection bias problem could occur if t-test or regression analysis were used. The results show that a credit guarantee program enhances the financial performance of beneficiary firms.

• Journal of Hydrospace Technology
• /
• v.1 no.1
• /
• pp.75-88
• /
• 1995

Auto-Pilot System uses heading angle information via the position sensor and the rudder device to control the ship's direction. Most of the control logics are composed of the state estimation and control algorithms assuming that the measurement device and the actuator have no fault except the measurement noise. But such asumptions could bring the danger in real situation. For example, if the heading angle measuring device is out of order the control action based on those false position information could bring serious safety problem. In this study, the control system including improved method for processing the position information is applied to the Auto-Pilot System. To show the difference between general state estimator and F.D.F., BJDFs for the sensor and the actuator failure detection are designed and the performance are tested. And it is shown that bias error in sensor could be detected by state-augmented estimator. So the residual confined in the 2-dimension in the presence of the sensor failure could be unidirectional in output space and bias sensor error is much easier to be detected.

• Survey Research
• /
• v.4 no.2
• /
• pp.63-86
• /
• 2003

The objective of this study was to investigate the existence of starting point bias in the bidding game contingent valuation elicitation technique when determining willingness to pay (WTP) for improving the qualify of running water in Seoul and its vicinity. Of all existing contingent techniques, the bidding game most closely mimics the normal price taking behavior in local markets. Three different starting points (low, medium and high) were used to determine WTP and the existence of starting point bias in the meanwhile respectively. The respondents were randomly assigned to the three different starting point groups in order to ensure homogeneity, so that any variation seen in WTP could be attributed to the starting point effects. And a pretested interviewer-administered questionnaire used to elicit WTP. Non-parametric test and the logit model were used to analyze the data for evidence of starting point bias. In this instance, the high starting point group had a high WTP, and low starting point group had a low WTP. This means there exist starting point bias in estimating WTP by bidding game in this instance. This finding might signal that people may actually be making up their minds on the maximum amount they are willing to pay for running water service as a bidding iterating is going on and is influenced by the starting point used in the bidding game. The problem of slaking point bias can be avoided if the respondent is asked directly for the maximum WTP without payment cards or a bidding game. But such a question is perceived as being very difficult to answer and this leads to problems of non-response and being unrealistic.

• Proceedings of the Korean Society of Near Infrared Spectroscopy Conference
• /
• 2001.06a
• /
• pp.4107-4107
• /
• 2001

Previous works have shown the viability of NIRS technology for the prediction of fatty acids in Iberian pig fat, but although the resulting equations showed high precision, in the predictions of new samples important fluctuations were detected, greater with the time passed from calibration development to NIRS analysis. This fact makes the use of NIRS calibrations in routine analysis difficult. Moreover, this problem only appears in products like fat, that show spectrums with very defined absorption peaks at some wavelengths. This circumstance causes a high sensibility to small changes of the instrument, which are not perceived with the normal checks. To avoid these inconveniences, the software WinISI 1.04 has a mathematic algorithm that consist of create a “Repeatability File”. This file is used during calibration development to minimize the variation sources that can affect the NIRS predictions. The objective of the current work is the evaluation of the use of a repeatability file in quantitative NIRS analysis of Iberian pig fat. A total of 188 samples of Iberian pig fat, produced by COVAP, were used. NIR data were recorded using a FOSS NIRSystems 6500 I spectrophotometer equipped with a spinning module. Samples were analysed by folded transmission, using two sample cells of 0.1mm pathlength and gold surface. High accuracy calibration equations were obtained, without and with repeatability file, to determine the content of six fatty acids: miristic (SECV$\sub$without/=0.07% r$^2$$\sub$without/=0.76 and SECV$\sub$with/=0.08% r$^2$$\sub$with/=0.65), Palmitic (SECV$\sub$without/=0.28 r$^2$$\sub$without/=0.97 and SECV$\sub$with/=0.24% r$^2$$\sub$with/=0.98), palmitoleic (SECV$\sub$without/=0.08 r$^2$$\sub$without/=0.94 and SECV$\sub$with/=0.09% r$^2$$\sub$with/=0.92), Stearic (SECV$\sub$without/=0.27 r$^2$$\sub$without/=0.97 and SECV$\sub$with/=0.29% r$^2$$\sub$with/=0.96), oleic (SECV$\sub$without/=0.20 r$^2$$\sub$without/=0.99 and SECV$\sub$with/=0.20% r$^2$$\sub$with/=0.99) and linoleic (SECV$\sub$without/=0.16 r$^2$$\sub$without/=0.98 and SECV$\sub$with/=0.16% r$^2$$\sub$with/=0.98). The use of a repeatability file like a tool to reduce the variation sources that can disturbed the prediction accuracy was very effective. Although in calibration results the differences are negligible, the effect caused by the repeatability file is appreciated mainly when are predicted new samples that are not in the calibration set and whose spectrum were recorded a long time after the equation development. In this case, bias values corresponding to fatty acids predictions were lower when the repeatability file was used: miristic (bias$\sub$without/=-0.05 and bias$\sub$with/=-0.04), Palmitic (bias$\sub$without/=-0.42 and bias$\sub$with/=-0.11), Palmitoleic (bias$\sub$without/=-0.03 and bias$\sub$with/=0.03), Stearic (bias$\sub$without/=0.47 and bias$\sub$with/=0.28), oleic (bias$\sub$without/=0.14 and bias$\sub$with/=-0.04) and linoleic (bias$\sub$without/=0.25 and bias$\sub$with/=-0.20).

• Journal of Korea Water Resources Association
• /
• v.50 no.1
• /
• pp.1-15
• /
• 2017

Information on radar rainfall with high spatio-temporal resolution over large areas has been used to mitigate climate-related disasters such as flash floods. On the other hand, a well-known problem associated with the radar rainfall using the Marshall-Palmer relationship is the underestimation. In this study, we develop a new bias correction scheme based on the quantile regression method. This study employed a bivariate copula function method for the joint simulation between radar and ground gauge rainfall data to better characterize the error distribution. The proposed quantile regression based bias corrected rainfall showed a good agreement with that of observed. Moreover, the results of our case studies suggest that the copula function approach was useful to functionalize the error distribution of radar rainfall in an effective way.

• The Korean Journal of Applied Statistics
• /
• v.21 no.3
• /
• pp.361-375
• /
• 2008

We propose proper methods to impute the item nonresponse in 4-8-4 rotation sample survey. We consider nonignorable nonresponse mechanism that can happen when survey deals with sensitive question (e.g. income, labor force). We utilize modeling imputation method based on Bayesian approach to avoid a boundary solution problem. We also estimate a interview time bias using imputed data and calculate cell expectation and marginal probability on fixed time after removing estimated bias. We compare the mean squared errors and bias between maximum likelihood method and Bayesian methods using simulation studies.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.12
• /
• pp.1595-1603
• /
• 2020

Online defamation incidents such as Internet news comments on portal sites, SNS, and community sites are increasing in recent years. Bias and hate expressions threaten online service users in various forms, such as invasion of privacy and personal attacks, and defamation issues. In the past few years, academia and industry have been approaching in various ways to solve this problem The purpose of this study is to build a dataset and experiment with deep learning classification modeling for detecting various bias expressions as well as hate expressions. The dataset was annotated 7 labels that 10 personnel cross-checked. In this study, each of the 7 classes in a dataset of about 137,111 Korean internet news comments is binary classified and analyzed through deep learning techniques. The Proposed technique used in this study is multi-channel CNN model with attention. As a result of the experiment, the weighted average f1 score was 70.32% of performance.