• Journal of the Korean Data and Information Science Society
• /
• 제25권5호
• /
• pp.1127-1135
• /
• 2014

Genome-wide association studies (GWAS) are designed to discover genetic variants such as single nucleotide polymorphisms (SNPs) that are associated with human complex traits. Although there is an increasing interest in the application of GWAS methodologies to population-based cohorts, many published GWAS have adopted a case-control design, which raise an issue related to a sampling bias of both case and control samples. Because of unequal selection probabilities between cases and controls, the samples are not representative of the population that they are purported to represent. Therefore, non-random sampling in case-control study can potentially lead to inconsistent and biased estimates of SNP-trait associations. In this paper, we proposed inverse-probability of sampling weights based on disease prevalence to eliminate a case-control sampling bias in estimation and testing for association between SNPs and quantitative traits. We apply the proposed method to a data from the Korea Association Resource project and show that the standard estimators applied to the weighted data yield unbiased estimates.

• 한국정보통신학회논문지
• /
• 제14권4호
• /
• pp.948-958
• /
• 2010

MRI 시스템이 수집하는 인체신호는 매우 미약하기 때문에 영상화 과정을 거치면서 외부 잡음이나 시스템 불안정성에 의한 영향을 쉽게 받을 수 있다. 따라서 본 논문에서는 저 자장 MRI시스템에서 RF 수신코일의 디자인적 요소에 의해 발생되는 불균일성을 분석하여 영상의 균일도 향상 기법을 제안한다. 본 논문에서는 MRI영상의 신호강도 불균일성을 보정하기 위한 방법 중에서 팬텀 데이터를 이용하여 확장된 크기를 갖는 3차원 bias 볼륨 데이터를 획득하기 위한 방법을 제안함으로써 다양한 크기를 갖는 영상의 보정이 가능하도록 하였다. 제안된 bias 데이터의 최적화 기법을 적용하여 실험을 수행한 결과 단일 bias 데이터의 사용으로 다양한 영상법에 의한 영상을 효과적으로 보정할 수 있음을 확인 하였다.

• Journal of Information Processing Systems
• /
• 제19권5호
• /
• pp.641-651
• /
• 2023

Malicious hate speech and gender bias comments are common in online communities, causing social problems in our society. Gender bias and hate speech detection has been investigated. However, it is difficult because there are diverse ways to express them in words. To solve this problem, we attempted to detect malicious comments in a Korean hate speech dataset constructed in 2020. We explored bidirectional encoder representations from transformers (BERT)-based deep learning models utilizing hyperparameter tuning, data sampling, and logits ensembles with a label distribution. We evaluated our model in Kaggle competitions for gender bias, general bias, and hate speech detection. For gender bias detection, an F1-score of 0.7711 was achieved using an ensemble of the Soongsil-BERT and KcELECTRA models. The general bias task included the gender bias task, and the ensemble model achieved the best F1-score of 0.7166.

• 한국경영과학회지
• /
• 제34권2호
• /
• pp.91-100
• /
• 2009

"How does advertising affect sales?" is the fundamental issue of modern advertising research. There is an interesting issue for estimating carryover effects of advertising on sales, and the aggregated data biases exist in the duration of advertising effect. This research suggests an extended model of Koyck Model which is employed for micro-data (Koyck 1954) to estimate aggregated advertising data, and empirically shows the aggregated data bias. Our developed model with the aggregated level of actual advertising data is more appropriate than the basic Koyck model for micro-data. The result figures out that it is important to consider the disaggregated data level in the analysis of dynamic effects of adverting such as carryover effects.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2000년도 추계종합학술대회 논문집(1)
• /
• pp.353-356
• /
• 2000

Multi-bias parameter extraction technique for HEMT small signa] equivalent circuits is presented in this paper. The technique in this paper uses S-parameters measured at various bias points in the active region to construct one optimization problem, of which the vector of unknowns contains only a set of bias-independent elements. Tests are peformed on measured S-parameters of a pHEMT at 30 bias points. Results indicate that the calculated S-parameters is similar to the measured data.

• Communications for Statistical Applications and Methods
• /
• 제11권3호
• /
• pp.553-562
• /
• 2004

In this short communication we discuss the bias problems of CART in split variable selection and suggest a method to reduce the variable selection bias. Penalties proportional to the number of categories or distinct values are applied to the splitting criteria of CART. The results of empirical comparisons show that the proposed modification of CART reduces the bias in variable selection.

• 대한물리치료과학회지
• /
• 제30권4호
• /
• pp.82-91
• /
• 2023

Background: Randomized controlled trials (RCTs) provide evidence on the effectiveness and safety of interventions and inform systematic reviews and guideline preparation for clinical application. However, methodological flaws can occur in many RCTs, and Cochrane's risk of bias version 2 (RoB2) can be used to evaluate RCTs' risk of bias (RoB). However, physical therapy RCTs in Korea did not confirm RoB. Therefore, the purpose of this study was to evaluate RoB using RoB2 in RCTs published in the Korean Physical Therapy Journal. Design: Review. Methods: The RCTs subject to evaluation were RCTs published in 11 physical therapy journals in Korea from 2018 to 2022. RoB2 evaluated a total of five domains: bias arising from the randomization process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in measurement of the outcome, and bias in selection of the reported result. Results: A total of 616 RCTs were evaluated. As for bias arising from the randomization process, high risk was the highest at 555 (90.1%), followed by low risk at 41 (6.7%) and some concerns at 20 (3.2%). For bias due to deviations from intended interventions, the proportion of some concerns was the highest at 390 (63.3%), followed by high risk at 218 (35.4%) and low risk at 8 (1.3%). As for the bias due to missing outcome data, the rate of low risk was the highest at 399 (64.8%), followed by high risk at 159 (25.8%) and some concerns at 58 (9.4%). As for bias in measurement of the outcome, high risk was the highest at 294 (47.7%), followed by low risk at 224 (36.4%) and some concerns at 98 (15.9%). In the bias due to missing outcome data, the ratio of high risk was the highest at 610 (99%), followed by low risk at 4 (0.7%) and some concerns at 2 (0.3%). Conclusion: Most of the RoB evaluation results of RCTs published in the Korean Physical Therapy Journal were rated as high risk. Methodological quality of RCTs needs to be improved.

• 정보교육학회논문지
• /
• 제26권4호
• /
• pp.273-284
• /
• 2022

본고의 목표는 데이터 편향 인식 교육에서 기계학습 플랫폼의 사용을 제안하는 것이다. 학습자들이 인공지능 데이터 및 시스템을 다루거나 인공지능윤리 요소 중 데이터 편향에 의한 피해를 방지하고자 할 때 인지할 수 있는 역량을 배양할 수 있다. 구체적으로, 머신러닝포키즈를 활용해 데이터편향 학습을 하는 방법을 AI야구심판 사례를 통해 제시한다. 학습자는 구체적 주제선정, 선행연구 검토, 기계학습 플랫폼에서 편향/비편향 데이터의 입력 및 테스트 데이터 구성, 기계학습의 결과 비교, 결과를 통해 얻을 수 있는 데이터 편향에 대한 함의를 제시한다. 이러한 과정을 통해서 학습자는 인공지능 데이터 편향이 최소화되어야 한다는 점과 데이터 수집 및 선정이 사회에 미치는 영향을 체험적으로 배울 수 있다. 이 학습방법은 문제기반의 자기주도 학습의 용이성, 코딩교육과의 결합가능성, 그리고 인문사회적 주제와 인공지능 리터러시와 결합을 추동한다는 의의를 가진다.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2001년도 ICCAS
• /
• pp.127.6-127
• /
• 2001

In this paper, we deal with two bias compensation algorithms of acceleration sensor for measuring the galloping on power transmission line. Firstly, the block diagram of galloping measurement system is given and a galloping model is presented. Secondly, two compensation algorithms, a simple compensation and a period compensation, are proposed. A simple compensation algorithm use the drafts of velocity and distance at fixed periods, so it is useful for constant bias case. Next, a period compensation algorithm can compensate a periodic bias. This algorithm use the previous measured data and compensated data for constant period, where the period is obtained by FFT method. Lastly, the effectiveness of proposed algorithms is verified by comparing between two algorithms in simulation, and its characteristics and the bias error bound are shown, respectively.

• 대한원격탐사학회지
• /
• 제34권6_1호
• /
• pp.997-1007
• /
• 2018

가뭄감시는 기후변화로 인해 빈번히 발생하는 자연재해를 저감하기 위해 필요한 중요한 요소 중의 하나이다. 한반도 지역의 가뭄감시를 수행하기 위해서는 위성기반 강수량을 관측하는 것이 필요하다. 본 연구에서는 위성기반의 원시위성강우자료와 편의보정한 위성자료를 이용하여 위성기반 강수량의 정확도를 확인하였다. 서로 다른 공간/시간 해상도를 가지는 원시위성자료(TRMM TMPA, GPM IMERG)를 10 km로 재격자화 하고, 일단위로 변환하였다. 최종적으로 원시위성강우의 표준 시간대를 한반도 표준시(GMT+9)로 변환하여 데이터베이스를 구축하였다. 한반도를 대상지역으로 선정하여, 지상관측자료와 검증을 실시하였다. 편의보정 기법은 GRA-IDW 기법을 선정하여 수행하였다. 먼저 원시위성자료를 검증한 결과를 살펴보면, 상관계수는 1998년부터 2017년까지 0.775로 비교적 정확도가 높게 나왔으며, TRMM TMPA, GPM IMERG 각각의 10 km 일강수량 상관계수값은 0.776, 0.753으로 크게 차이 나지 않았다. BIAS값은 원시위성자료 값이 지상관측자료보다 과대추정하는 것으로 나타났다. 편의보정한 위성자료를 검증한 결과를 살펴보면, 상관계수와 RMSE가 편의보정 전보다 개선된 값을 보여주고 있다. 본 연구에서 검증한 위성강우자료는 가뭄감시시스템의 기초자료로 충분히 활용할 수 있으며, 향후 미계측지역의 가뭄관리 의사결정을 위한 격자자료로 활용할 수 있을 것으로 판단된다.