Earticle

본 논문은 공항 X-ray 보안 검색의 정확도를 높이기 위해, 현장 보안검색요원의 밀도 중심 판독 기법을 모사하여 이미지의 시각적 특징뿐만 아니라 물체의 물리적 속성인 '추정 밀도' 정보를 함께 활용하는 다중 입력(Multi-input) 기반의 딥러닝 모델 을 제안한다. 최근 지능화되는 위해물품 반입 시도에 대응하기 위해 정밀한 검색 시스템이 요구되고 있으나, 기존의 단일 시 각 정보 의존 방식은 형태가 유사한 이종 물질 식별에 한계가 있다. 이를 해결하기 위해 본 연구는 CNN 기반의 VGG16과 EfficientNetB0 모델에 그레이스케일 강도에서 추출한 밀도 정보를 추가 입력 변수로 결합하여 분류 성능을 강화하였다. 특히, 현직 항공보안검색요원의 자문을 통해 선별된 11종의 핵심 위해물품 데이터를 활용하여 실험의 현실성을 확보하였다. 실험 결 과, 제안된 방식은 기존 모델 대비 오분류율을 유의미하게 감소시켰으며, 특히 중첩되거나 시각적으로 구분이 어려운 객체 식 별에 있어 밀도 정보가 중요한 판단 근거로 작용함을 입증하였다.

This paper proposes a multi-input deep learning approach that mimics the density-centric screening technique of aviation security officers by combining visual features with "estimated density" information derived from physical properties. As attempts to smuggle hazardous items become more sophisticated, existing systems relying solely on visual information face limitations in distinguishing materials with similar shapes. To address this, we designed a method that integrates density features extracted from grayscale intensity into CNN-based models (VGG16 and EfficientNetB0) to enhance classification performance. In particular, the reliability of the dataset was secured by selecting 11 core hazardous item classes through consultation with active aviation security officers. Experimental results demonstrate that the proposed multi-input method significantly reduces misclassification rates compared to baseline models and proves the effectiveness of density information, especially in identifying overlapped or visually ambiguous objects.

이 연구는 2005년부터 2009년 기간 동안에 한국과 중국의 은행들의 효율성을 측정하고 두 국가 은행들의 효율성에 대한 정책적인 의미를 찾아 보고자 하였다. 주요 발견사실은 다음과 같다. (1) 한국의 은행들은 지난 5냔 동안 규모수익불변(CRS)나 규모수익증가(IRS)의 경향이 많았고 중국의 운행들은 규모수익감소(DRS)의 경향이 많았다. (2) 중국상인은행(CMB)과 상해푸동개발은행(SPDB)은 네 개의 중국의 대형 국책은행의 효율성보다 높았다. (3) 신한은행의 효율성이 한국의 은행 중 생산성이 가장 높았고 ,한국의 은행의 효율성이 중국의 은행의 효율성보다 전반적으로 높게 나타났다. 그러므로 중국의 대형국책은행들은 외국은행이나 중국의 다른 상업은행기관들과의 경쟁에 대처할 수 있는 효율성 증대에 많은 노력이 경주되어야 한다고 생각된다.

This study used the Data Envelopment Analysis (DEA) to measure the efficiency of Chinese and Korean banks from 2005 to 2009, and explored the implications of the banks efficiency in the two countries. Main findings of this paper are as follows. (1) Korean banks have had CRS and IRS trends during these 5 years, but most Chinese banks showed DRS trend. (2) The efficiency of CMB and SPDB in China is higher than Big Four State-owned Banks. (3) The efficiency of SB is the highest in Korea. And the results showed that the banks of Korea have relatively higher efficiency than that of China. Therefore, Chinese banks, expecially the Big Four State-owned Banks, must endeavor to improve the efficiency to cope with the competition from the other Chinese joint-stock commercial banks and foreign banks.

본고는 다양한 머신러닝 방법과 폭넓은 설명변수를 활용하여 한국 주식시장의 중요 변동성 지표인 VKOSPI 지수를 예측/분석하였다. 일반적인 금융시장 지표뿐만 아니라 주식시장 자금흐름 및 수급동향과 관련된 변수들도 설명변수로 이용하였으며, 특히 변동성 예측을 위해 새로운 Multi-Input LSTM(Long Short-Term Memory) 모형을 제안하고 예측력을 평가하였다. 표준적인 변동성 모형인 HAR(Heterogeneous Autoregressive) 모형에 비해 머신러닝 모형들의 예측력이 전반적으로 우수한 것으로 나타났는데, 예측 대상기간이 짧은 경우 선형 모형이 그리고 예측 대상기간이 긴 경우 비선형 모형이 우수한 예측력을 보였다. 또한 Multi-Input LSTM 모형이 기존 LSTM 모형보다 예측력이 우수하다는 점과 주식시장 자금흐름 및 수급동향 관련 변수들이 Multi-Input LSTM 모형에서 중요한 역할을 한다는 것을 확인하였다.

This study explores the forecasting of the VKOSPI index, a vital volatility indicator for the Korean stock market, using advanced machine learning models. VKOSPI, similar to the VIX index in the U.S., reflects market expectations of future volatility and is crucial for risk management and financial derivatives pricing. Traditional forecasting models, such as the Heterogeneous Autoregressive (HAR) model, have been widely used in this domain. However, this research introduces a novel Multi-Input Long Short-Term Memory (LSTM) model, which incorporates a wide range of explanatory variables, including general financial variables, stock market fund flows and stock market transaction trend, to improve predictive accuracy. The Multi-Input LSTM model is designed to sequentially process categorized variables through multiple input layers, allowing it to capture complex patterns that are missed by simpler models. The study, based on data from January 2016 to March 2023, compares the forecasting performance of the Multi-Input LSTM with various other machine learning models, such as Random Forest, XGBoost, and linear models like LASSO and Elastic Net. With a total of 216 explanatory variables, including lagged values, the models are tested across different forecast horizons, including 1 day, 5 days, 10 days, and 22days. The results show that machine learning models, particularly non-linear ones such as LSTM, Random Forest, and XGBoost, outperform the HAR model, especially for longer forecast periods. The Multi-Input LSTM demonstrates superior accuracy, particularly when the variables about stock market transaction and stock market fund flow are included, highlighting their importance in predicting volatility. Furthermore, the study uses rigorous statistical tests to validate the enhanced performance of the Multi-Input LSTM model over traditional approaches. The study contributes to the growing field of machine learning in finance, demonstrating that advanced models like the Multi-Input LSTM can effectively handle the complexities of financial time series data.

This research proposes a new advanced control method and demonstrates its realization in part. By incorporating shape machining and cutting force control at a time, this integrated scheme makes it possible to machine a desired shape and avoid the trouble of programming feedrate and spindle speed before machining and also reduce the shape error. The main idea proposed to achieve those goals consists in giving commanded path and desired cutting force at the same time. which makes it possible for position and force controller to distribute the corresponding velocity of individual axes and main spindle by an appropriate interpolation. That indicates we can replace the built-in interpolator of commercial machine tools by the developed algorithm.

Most of the data envelopment analysis (DEA) models evaluate the relative efficiency of a decision making unit (DMU) based on the assumption that inputs in a specific period are consumed to produce the output in the same period of time. However, there may be some time lag between the consumption of input resources and the production of outputs. A few models to handle the concept of the time lag effect have been proposed. This paper suggests a new multi-period input DEA model considering the consistent time lag effects. Consistency of time lag effect means that the time delay for the same input factor or output factor are consistent throughout the periods. It is more realistic than the time lag effect for the same output or input factor can vary over the periods. The suggested model is an output-oriented model in order to adopt the consistent time lag effect. We analyze the results of the suggested model and the existing multi period input model with a sample data set from a long-term national research and development program in Korea. We show that the suggested model may have the better discrimination power than existing model while the ranking of DMUs is not different by two nonparametric tests.

이 논문에서는 국제무역 및 제품에 내재된 이산화탄소 배출량을 계산하기 위하여 환경 산업연관분석모형을 이용하였다. 분석결과, 첫째, 탄소 배출집약도 분석결과, 전력, 비금속, 철강, 비철금속, 항공수송, 육상/해상수송의 탄소배출 집약도가 높게 나타났다. 둘째, 우리나라는 호주/뉴질랜드, 아르헨티나, 브라질, 인도네시아, 일본, 러시아, 중동국가, 남아공, 중위소득국가 등과 교역할때 제품에 내재된 탄소를 수입하고, 캐나다, 프랑스, 독일, 인도, 이탈리아, 멕시코, 터키, 영국, 미국, 기타유럽 등과 교역 시 탄소를 제품에 내재화하여 수출하는 것으로 분석되었다. 셋째, 주요 국가별 환경수지를 분석한 결과, 중국, 러시아, 중동국가 등 개도국이 환경수지 흑자를 기록하는 반면 OECD 국가들은 환경수지 적자인 것으로 분석되었다. 이 논문은 소비측면의 탄소배출을 분석함으로써 국제무역에서 발생하는 환경수지를 파악할 수 있다.

This paper aims to analyze the embodied carbon trade using MRIO(Multi-Regional Input-Output) model. A Multi-Regional Input-Output model has been chosen as the framework to be used for accounting CO2 emissions embodied in the commodity bundle. The data base of the Global Trade Analysis Project (GTAP) was used to construct the MRIO model with full linkages. The main results are as follows. Firstly, the embodied carbon is dispersed both across commodities and regions. Generally, chemicals, ferrous and non-ferrous metals and transportation services tend to be both energy and carbon intensive. Secondly, the most carbon-intensive commodities are electricity, metal and mineral products, transportation services and refined energy products. Thirdly, We decompose the embodied carbon of goods into direct and indirect factors. The electricity and transportation sectors are more carbon-intensive in production directly than indirect domestic and indirect imported. Fourthly, the Korea exports more carbon in trade with Russia, Japan, China, India in iron and steel sector. Fifthly, the Korea exports more carbon in trade with developed countries,but imports more carbon in trade with developing countries. Finally, the United States,United Kingdom, Germany, and Japan import more carbon in trade with non-OECD states.

This paper investigates regional economics effects of the 2006 Busan Biennale using a multi-regional input-output(MRIO) model. Busan Biennale, as a cultural event, is an ambitious project to develop Busan to be an advanced international and worldwide artistic city. The result of this analysis is as follows: The result of the economic analysis reveals that the efficiency of the output against the investment is proved very high. The event could generate 41,223 million won of output, 19,794 million won of added value and 873 persons of employment. It shows that cultural projects can make a successful contribution to the economic development of a city. It is expected that this study will be a helpful case for similar studies and at the same time a reference resource for policy analysis concerned. Especially it will help those who consider the international event as non-productive and consumptive correct their prejudices. From these results, it is desirable that the venue of Busan Biennale be expanded to wider areas of southern parts of Korea including Ulsan, Kyung-Nam, and so on.

The purposes of this study are (a) to analyze the economic impacts of the 2002 Korea-Japan World Cup on regional economies, using the multi-regional input-out (MRIO) model and (b) to compare the study results between earlier related researches (e.g. those of KDI or others). The 9 region input-output table is derived from 402 transactions tables of Bank of Korea and finally aggregated into 22 industries. The results of this study showed that construction investments of World Cup stadiums in 10 cities and total World Cup tourism receipts from lodging, food and beverage, transportation, and shopping and entertainment industries could generate 6,520.2 billion won of output, and 3,037.5 billion won of added value. The results of this research also revealed that new employment opportunities of 91,700 were expected all around the country. The detailed economic effects of 2002 World Cup are classified into 9 regions in term of output, value added, and employment. Implications of this study are discussed and suggestions are developed for future research.