Earticle

이 글은 모든 일하는 사람의 노동권을 실현하기 위한 방법으로서, 두 가지 제안을 하는데 집중하였다. 첫째는 근로기준법상 근로자 개념의 불확정성으로 인하 여 발생하는 오분류 문제를 완화하기 위한 최소한의 조치로서 근로기준법상 근로 자 개념을 원용하고 있지만 반드시 그러할 당위성은 없는 근로자퇴직급여보장법, 최저임금법 등의 근로자 개념을 해당 법의 목적과 취지에 따라 새롭게 재구성하 는 한편 고용관계추정 규정을 두는 것, 그리고 근로기준법에서 근로자추정규정을 두는 것이다. 둘째는 지난 5년여간 꾸준하게 제안되었던 일하는 사람을 위한 노 동기본법 제정을 다시 촉구하면서, 기본법으로서의 의미와 노동기본법 안에 담겨 야 할 노동기본권을 확인한 것이다. 그리고 여기에는 차별을 받지 않을 권리, 투 명한 계약을 체결할 권리, 일터의 안전과 건강에 대한 권리, 쉴 권리, 스스로를 대표할 권리, 일･가정양립을 위한 권리, 사회보험의 적용을 받을 권리, 정보접근 과 정보정정의 권리, 사업주의 설명 책임 등을 포함해야 한다고 하였다. 이외 무 수한 권리들이 떠오를 수 있지만, 노동기본법에 담겨야 할 가장 기본적인 노동권 이라고 생각한 것들이다. 이러한 노동권은 특히 비전형노동자의 노동 수행 방법 과 종류를 고려한 것이기도 하고, 근로기준법상 근로자와는 달리 사업장이라는 장소에 기반을 두지 않고 노동을 수행해야 하는 특성을 반영한 것이기도 하다. 이것이 반드시 유일한 정답이라고 말할 수는 없다. 그러나 아직 가보지 않은 길을 두고 틀렸다고 말할 수도 없다. 근래에는 일하는 사람을 위한 노동기본법 외양을 갖춘 듯하면서도 그 내용은 사실상 제3의 노동자 군을 조형함으로써 근 로자와 근로자 아닌 자를 구분하고 노동권이 아닌 수익권의 형태로 비전형노동 자들의 권리보호를 말하는 경우도 있는 듯하지만, 일하는 사람을 위한 노동기본 법은 근로기준법상 근로자에 대한 오분류 문제를 최소화하면서 모든 노동자의 노동권을 보호하는데 그 목적을 두지 않을 수 없을 것이다.

This article focuses on two proposals as a way of realising labour rights for all workers. The first is to reconstruct the concept of worker and presume worker status as a minimum measure to mitigate the misclassification problem caused by the uncertainty of the concept of ‘worker’ under the Labour Standards Act. This includes reconstructing the concept of employee in labour laws such as the Employee Retirement Income Security Act and the Minimum Wage Act, which are based on, but not necessarily the same as, the concept of employee in the Labour Standards Act. Second, it reiterated the call for the enactment of the Basic Labour Law for Working People, which has been proposed for the past five years, and confirmed its importance as a basic law and the basic labour rights that should be included in it. This basic law should include the right not to be discriminated against, the right to transparent contracts, the right to safety and health at work, the right to rest, the right to self-representation, the right to reconcile work and family life, the right to be covered by social security, the right to access and correct information and the employer's responsibility to explain. There may be countless other rights, but these are the most basic labour rights that I believe should be included in a basic labour law. These labour rights are specifically designed to take into account the methods and types of work performed by non-traditional workers, and also to reflect the fact that, unlike workers covered by the Labour Standards Act, they are not based at a place of business. It is not necessarily the only answer. But it's not wrong either. In recent years, there have been a number of laws that give the appearance of a basic labour law for working peolpe, but in fact create a third group of workers, distinguish between employees and non-employees, and protect the rights of non-traditional workers in the form of profit rights rather than labour rights. However, the purpose of the basic labour law for workers is to protect the labour rights of all workers while minimising the problem of misclassification of workers under the Labour Standards Act.

딥뉴럴네트워크는 머신러닝 분야 중 이미지 인식, 사물 인식 등에 좋은 성능을 보여주고 있다. 그러나 딥뉴럴네트워크는 적 대적 샘플(Adversarial example)에 취약점이 있다. 적대적 샘플은 원본 샘플에 최소한의 noise를 넣어서 딥뉴럴네트워크가 잘 못 인식하게 하는 샘플이다. 그러나 이러한 적대적 샘플은 원본 샘플간의 최소한의 noise을 주면서 동시에 딥뉴럴네트워크가 잘못 인식하도록 하는 샘플을 생성하는 데 시간이 많이 걸린다는 단점이 있다. 따라서 어떠한 경우에 최소한의 noise가 아니 더라도 신속하게 딥뉴럴네트워크가 잘못 인식하도록 하는 공격이 필요할 수 있다. 이 논문에서, 우리는 신속하게 딥뉴럴네트 워크를 공격하는 것에 우선순위를 둔 신속한 오인식 샘플 생성 공격을 제안하고자 한다. 이 제안방법은 원본 샘플에 대한 왜 곡을 고려하지 않고 딥뉴럴네트워크의 오인식에 중점을 둔 noise를 추가하는 방식이다. 따라서 이 방법은 기존방법과 달리 별 도의 원본 샘플에 대한 왜곡을 고려하지 않기 때문에 기존방법보다 생성속도가 빠른 장점이 있다. 실험데이터로는 MNIST와 CIFAR10를 사용하였으며 머신러닝 라이브러리로 Tensorflow를 사용하였다. 실험결과에서, 제안한 오인식 샘플은 기존방법에 비해서 MNIST와 CIFAR10에서 각각 50%, 80% 감소된 반복횟수이면서 100% 공격률을 가진다.

Deep neural networks (DNNs) provide good performance for machine learning tasks such as image recognition and object recognition. However, DNNs are vulnerable to an adversarial example. An adversarial example is an attack sample that causes the neural network to recognize it incorrectly by adding minimal noise to the original sample. However, the disadvantage is that it takes a long time to generate such an adversarial example. Therefore, in some cases, an attack may be necessary that quickly causes the neural network to recognize it incorrectly. In this paper, we propose a fast misclassification sample that can rapidly attack neural networks. The proposed method does not consider the distortion of the original sample when adding noise. We used MNIST and CIFAR10 as experimental data and Tensorflow as a machine learning library. Experimental results show that the fast misclassification sample generated by the proposed method can be generated with 50% and 80% reduced number of iterations for MNIST and CIFAR10, respectively, compared to the conventional Carlini method, and has 100% attack rate.

심리학의 모든 분야에 걸쳐서 능력, 성격특질, 혹은 태도, 가치관 등이 대비되는 집단들을 구성하고, 그러한 집단들이 특정한 구성개념(construct) 상에서 어떤 차이를 가지는지를 비교검증하는 연구 및 분석설계가 많이 활용되고 있다. 그런데 집단 구분 혹은 독립변인의 조작을 위하여 쓰이는 측정도구의 신뢰도가 완벽하지 않은 한, 그러한 측정도구를 이용하여 집단구분을 하는 경우 반드시 할당오류 혹은 분류오류를 범하게 된다. 이 분류오류는 말하자면 독립변인의 조작에서의 오류(error)이다. 반면에, 이러한 연구 및 분석설계를 위해서 자주 활용되는 분석기법들인 변량분석이나 희귀분석은 독립변인이 오차를 가지지 않는다는 가정을 한다. 측정도구를 이용한 독립변인의 조직에서 분류오류를 완전히 없애는 것은 불가능하지만, 특정한 척도를 집단분류를 위하여 이용하는 경우 분류오류가 어느 정도 발생할 것인지를 파악하는 것은 결과의 일반화와 해석의 일관성을 위하여 중요한 정보가 된다. 본 논문은 심리척도상의 연속적인 점수에 기초하여 집단을 분류할 경우 분류 오류의 확률이 얼마인지 파악하는 방법과 절차를 항목반응이론(item response theory)의 개념적 틀안에서 실제 자료를 예로 들어 기술한 것이다. Hulin, Drasgow, and Parsons (1983) 가 기술한 방법에 의해 추정된 분류오류확률의 해석에 있어서 주의할 점이 논의되었다.

The purpose of the present paper was to demonstrate a method of estimating the probability of misclcissification when a Scale is used to classify respondents arbitrarily into the high and low groups based upon the scores on the scale. Determination of misclassification probability is made possible because Item Response Theory provides theoretically sound estimation of latent trait. In this study, misclassification was defined by the hypothetical incidents in which the subjects with the actual degree of traits below the median are assigned into the high group and the subjects with the actual degree of traits above the median are assigned into the low group based upon the scores on the scale. The estimated misclassification probability associated with Beck Hopelessness Scale was .06. The interpretive meaning of the misclassification probability as defined in this study was discussed. Finally cut-off pointe on the raw score scale reflecting those probabilities were provided for practical consideration, and important issues for further studies were suggested.

Background: Misreporting self-reported family history may lead to biased estimations. We used Bayesian methods to adjust for exposure misclassification. Materials and Methods: A hospital-based case-control study was used to identify breast cancer risk factors among Iranian women. Three models were jointly considered; an outcome, an exposure and a measurement model. All models were fitted using Bayesian methods, run to achieve convergence. Results: Bayesian analysis in the model without misclassification showed that the odds ratios for the relationship between breast cancer and a family history in different prior distributions were 2.98 (95% CRI: 2.41, 3.71), 2.57 (95% CRI: 1.95, 3.41) and 2.53 (95% CRI: 1.93, 3.31). In the misclassified model, adjusted odds ratios for misclassification in the different situations were 2.64 (95% CRI: 2.02, 3.47), 2.64 (95% CRI: 2.02, 3.46), 1.60 (95% CRI: 1.07, 2.38), 1.61 (95% CRI: 1.07, 2.40), 1.57 (95% CRI: 1.05, 2.35), 1.58 (95% CRI: 1.06, 2.34) and 1.57 (95% CRI: 1.06, 2.33). Conclusions: It was concluded that self-reported family history may be misclassified in different scenarios. Due to the lack of validation studies in Iran, more attention to this matter in future research is suggested, especially while obtaining results in accordance with sensitivity and specificity values.

A misclassified size-biased modified power series distribution (MSBMPSD) where some of the observations corresponding to x = c + 1 are misclassified as x = c with probability $\alpha$, is defined. We obtain its recurrence relations among the raw moments, the central moments and the factorial moments. Discussion of the effect of the misclassification on the variance is considered. To illustrate the situation under consideration some of its particular cases like the size-biased generalized negative binomial (SBGNB), the size-biased generalized Poisson (SBGP) and sizebiased Borel distributions are included. Finally, an example is presented for the size-biased generalized Poisson distribution to illustrate the results.

Background/Aims: Within metabolic dysfunction and alcohol-associated liver disease (Met- ALD), there exists a continuum where the condition can conceptually shift between being metabolic dysfunction-associated steatotic liver disease (MASLD) and alcoholic liver disease. However, alcohol use disorder (AUD) can be included in these diagnoses. The aim of this study was to investigate the prevalence and clinical characteristics of misclassified AUD among patients with MASLD and MetALD. Methods: The study included a total of 3,362,552 participants from the 2011 to 2012 National Health Screening Program. Steatotic liver disease was defined as having a hepatic steatosis index score of 36 or higher. Significant alcohol intake was calculated on the basis of self-report questionnaire responses. AUD was defined as having received medical care for an alcohol-related condition at least once during the study period. The mean follow-up period for participants was 9.8 years. Results: MASLD and MetALD prevalence were 23.8% and 1.9%, respectively. AUD was identified in 1.1% (8,481 individuals) of MASLD and 4.7% (2,989 individuals) of MetALD cases. Misclassified AUD was associated with significantly higher all-cause and liver-related mortality. Adjusted hazard ratios for liver-related mortality were 6.53 for AUD misclassified as MASLD and 6.98 for AUD misclassified as MetALD. Extrahepatic cancer mortality risk was also elevated (adjusted hazard ratio: 1.33 in MASLD and 1.44 in MetALD). Conclusions: A significant number of AUD cases were misclassified as MASLD and MetALD in cross-sectional assessment of alcohol consumption. Patients with AUD misclassified as MASLD or MetALD had higher liver-related mortality than the pure MASLD and MetALD groups.

The lack of a standardized classification system poses challenges for accurately assessing ultra-processed food (UPF) consumption in Korea. This scoping review compared the UPF consumption levels reported in Korean studies and analyzed the discrepancies to highlight the need for a unified UPF classification system tailored to Korea. Four online databases were searched to identify studies conducted in Korea. From 147 papers, 20 papers that examined UPFs as the main variables were selected and reviewed. Their consumption levels are reported. Ninety percent of the papers were cross-sectional studies, while 10% were prospective cohort studies. Most studies (72%) measured UPF consumption using the 24-h dietary recall (24HR), while 28% used food frequency questionnaires (FFQs). The most popular topic (65%) was the association between UPF consumption and health outcomes, with obesity-related outcomes being examined most frequently. A wide range of UPF consumption was observed: from 4.9% to 32.8% of total energy intake. The estimates from food-based FFQs were the lowest compared to those from the dish-based FFQs or 24HR. Significant variations in UPF consumption levels were observed across dietary assessment methods, researchers, and data sources. This scoping review highlights the need for an objective and standardized UPF classification system, developed through collaboration among researchers, to minimize the potential misclassification issues when estimating UPF consumption or examining its associations with the health outcomes in Korea. Establishing a "UPF Working Group" could serve as an effective starting point for this initiative, and it is expected to attract participation from more researchers interested in UPF studies.

Background: Misclassification of study designs of journals can hinder the readers from assessing the strengths and weaknesses of the study and evaluating the applicability of the study in the real-world setting. However, it seems that it is common for authors to neglect to classify the study design. Objective: We aimed to evaluate the accuracy of the classification of study designs and examine the common errors. Methods: This descriptive study analyzed four Asian dermatology journals listed in the science citation index expanded from January 2018 to December 2018. We investigated discrepancies between author-reported and actual study designs. Design Algorithm for Medical Literature on Intervention (DAMI) was used to determine the actual study design. Results: Of the 177 papers analyzed, only 72 articles (40.7%) revealed their study design and among them, 23 articles (32.0%) showed discrepancies between the author-reported and the actual study designs. Case-control studies were the most commonly misclassified study designs by authors. Conclusion: There were considerable differences between the author-reported study design and the actual study design in Asian dermatology journals. Proper classification of study designs by the authors is essential to strengthen evidencebased medicine.

The local behaviour of the surface formed by the perturbed maximum likelihood estimator of the squared Mahalanobis distance is investigated. The study of the local behaviour allows a simultaneous perturbation on the samples of interest and it is effective in identifying influential observations.

The influence of observations on the misclassification probability in multiple discriminant analysis under the equal covariance assumption is investigated by the local influence method. Under an appropriate perturbation we can get information about influential observations and outliers by studying the curvatures and the associated direction vectors of the perturbation-formed surface of the misclassification probability. We show that the influence function method gives essentially the same information as the direction vector of the maximum slope. An illustrative example is given for the effectiveness of the local influence method.

In the analysis of categorical data subject to misclassification errors, the observed cell proportions are adjusted by a misclassification probabilities and estimates of variances are adjusted accordingly. In this case, it is important to determine the extent to which misclassification probabilities are homogeneous within a population. This paper considers methods to evaluate the power of chi-squared tests for homogeneity with complex survey data subject to misclassification errors. Two cases are considered: adjustment with homogeneous misclassification probabilities; adjustment with heterogeneous misclassification probabilities. To estimate misclassification probabilities, logistic regression method is considered.

Confidence intervals for the population proportion in a double sampling scheme subject to false-positive misclassification are considered. The confidence intervals are obtained by applying Agresti and Coull's approach, so-called "adding two-failures and two successes". They are compared in terms of coverage probabilities and expected widths with the Wald interval and the confidence interval given by Boese et al. (2006). The latter one is a test-based confidence interval and is known to have good properties. It is shown that the Agresti and Coull's approach provides a relatively simple but effective confidence interval.

본 연구는 2009년부터 2013년 까지 H사가 수행한 건설 프로젝트의 회계처리를 중심으로, 도급공사와 자체사업의 자의적 회계처리 문제를 분석하고 그 시사점을 도출하였다. H사는 실질적으로 시행사로서의 통제권을 보유하고 있었음에도 불구하고, 해당 사업을 외부 도급공사로 회계처리하여 매출을 조기에 인식함으로써 수익인식의 시기 및 금액에 중대한 왜곡을 초래하였다. 본 사례는 회계기준의 해석 가능성과 산업 구조의 특수성이 회계오류 또는 회계부정으로 이어질 수 있음을 보여준다. 본 연구는 도급공사와 자체사업의 구분 기준 명확화, 외부감사 제도 개선, 공시 확대를 통한 투자자 보호, 회계 감리 강화 등 정책적 개선 방안을 제시함으로써, 회계투명성 확보와 시장신뢰 회복을 위한 실질적인 시사점을 제공하는데 공헌성이 있을 것이다.

This study examines the accounting errors and potential manipulations in the case of H firm from 2009 to 2013, focusing on the arbitrary classification between contract construction (turnkey projects) and self-developed projects. Although H firm effectively held control over certain projects, it reported them as third-party contracts, thereby recognizing revenue prematurely and overstating earnings. This practice deviated from the substance-over-form principle under K-IFRS and reflects broader systemic incentives, such as inflated performance evaluations through government-issued construction capability assessments. The case highlights how the discretionary interpretation of accounting standards and the structural nature of the construction industry can lead to financial reporting distortions. The study provides policy recommendations, including clearer criteria for revenue recognition, strengthened external audit quality, enhanced disclosures, and improved regulatory oversight. These findings underscore the need for comprehensive reforms to restore accounting transparency and investor trust in the construction sector.

.

This study analyzes the customs duty self-assessment system under the current customs law in Korea, with an emphasis placed on the analysis of customs classification, additional tax system, and the custom brokers's responsibility derived from error declaration of commodity classification. This paper also analyzes the roles and responsibilities of customs brokers and the importance of consignor's request in the customs clearing process for imports. Furthermore, we present consignor's cooperation plan with customs brokers in order to prevent declaration errors in the import process. One of the contribution of the current study is that we propose and establish a theory about the customs duty self-assessment system of Korea, additional tax system, and the custom broker's operation and responsibility as prescribed in the Korean law. The results of this study may be used as useful resources to recognize many registration points, in reference to the examples by conducting concrete analysis of precedent cases.

공공사업이나 비시장재의 편익 추정기법으로 주로 사용되는 가상가치평가법(CVM)의 기본 가정은 응답자들이 본인의 선호를 확실하게 알지만 연구자는 선호에 영향을 미치는 일부 변수들만 관측함으로써 모형에 확률적 부분이 존재한다는 것이다. 그러나 평가 대상이 일상 생활에서 자주 접하는 경우가 아니면 비록 설문조사의 조사지에서 상세하게 설명하더라도 응답자들은 본인의 선호를 확실하게 파악하지 못해서 잘못 응답할 수 있다. CVM모형에서 자주 이용되는 로짓모형과 같은 비선형모형에서는 종속변수의 측정오차는 추정 편의를 발생시킨다. 본 논문은 응답오류의 가능성을 모형에서 반영하기 위해서 Hausman et al. (1998)의 오분류 이항종속변수모형을 CVM에 적용하고, 적용된 모형의 응용사례로서 농업환경자원 변동 조사 사업의 편익분석을 고려한다. 분석결과, 응답오류 확률이 통계적으로 유의하며, 응답오류를 고려하지 않는 모형과 응답오류를 고려한 모형 간에 추정계수값 차이가 존재하고, 전자의 경우 계수들이 추정 편의를 갖는 것으로 나타났다. 응답오류를 고려한 CVM을 적용한 결과, 농업환경자원 변동 조사 사업의 가구당 편익은 3,831원으로 추정되었다.

This study considered the issue of response error in the Contingent Valuation Method(CVM), which is the primary approach used to assess public projects or nonmarket goods. The basic assumption of the CVM is that survey respondents recruited from the general public know their preferences with certainty. However, even when respondents are provided with detailed project explanations, this may not actually be the case, inducing incorrect responses and the measurement error of the dependent variable. In nonlinear models such as the logit model, which is frequently used in CVM method, this measurement error can cause estimation bias. To investigate this issue, this study used an existing misclassification binomial dependent variable method to adjust the logit model for the CVM (Hausman et al., 1998). This adjusted model was used to evaluate the Agricultural Environmental Change Survey project. In sum, the analysis produced the following results: (1) the response error probability was statistically significant, (2) differences existed between the two employed models (i.e., with and without considering response error), and (3) the estimated coefficients were biased in the model that considered response error. From the adjusted model, the benefit of the project was estimated as 3,831 won per household.

화장품 업계에서 온라인 리뷰 데이터는 급증하고 있으나 대부분이 긍정적이어서, 소수의 불만 리뷰를 효율적으로 분석하여 불만 요소를 파악하는 것이 기업에게 매우 중요하다. 본 연구는 한율 브랜드의 ‘어린쑥 수분진정 크림’을 대상으로 세 개의 온라인 플랫폼에서 수집한 총 22,119건의 리뷰 데이터를 이용하여 소비자 불만을 체계적으로 탐지하고 감성 변화 패턴을 분석하였다. 별점 기반과 텍스트 기반의 2단계 감성 분석을 수행하고, 한국어 언어학적 특성을 고려한 예외 처리 패턴을 정의하여 분석 정확성을 향상시켰다. Delta TF-IDF 기법으로 부정 리뷰 특화 키워드를 식별하고, LDA 토픽 모델링으로 불만 유형을 분류하였으며, 시계열 분석을 통해 감성 변화 추이를 파악하였다. 연구결과, 전체 리뷰의 97.8%가 긍정적이었으나 ‘그냥’, ‘아직’, ‘조금’ 등 부정 특화 키워드를 추출하였고, LDA 토픽 모델링을 통해 보습력 부족 및 사용감 불만, 피부 자극 및 성분 민감 반응, 리뉴얼 후 제형 변화 및 가성비 불만의 3개 주요 불만 토픽을 도출하였다. 시계열 분석에서는 리뷰 급증이 주로 긍정적 마케팅 활동과 시기적으로 일치하는 것으로 추정되며, 부정 리뷰 비율이 점차 안정화되는 패턴을 발견하였다. 본 연구의 다층적 분석 체계는 제품 개발, 마케팅 전략 수립, 실시간 고객 피드백 분석에 실용적 가치를 제공할 것으로 기대된다.

In the cosmetics industry, online review data is rapidly increasing, but since most reviews are positive, efficiently identifying complaint reviews has become an important challenge. This study systematically detected consumer complaints using 22,119 review data for HANYUL’s ‘Young Mugwort Moisture Soothing Cream’ from three online platforms. The research employed two-stage sentiment analysis with Korean language exception handling, Delta TF-IDF for negative-specific keywords, LDA topic modeling for complaint classification, and time series analysis for sentiment trends. Results showed 97.8% positive reviews, yet successfully extracted negative-specific keywords like ‘just’, ‘still’, and ‘a little’. Three major complaint topics were identified through LDA topic modeling: moisture insufficiency and texture dissatisfaction, skin irritation and ingredient sensitivity, and post-renewal complaints and value concerns. Time series analysis suggested review surges were temporally associated with positive marketing activities and gradually stabilizing negative review ratios. The multi-layered analytical framework provides practical value for product development, marketing strategy, and real-time customer feedback analysis.