기계학습알고리즘을 이용한 위험회복지수의 개발과 활용

김선웅 한국정보기술응용학회 2016 Journal of information technology applications & m Vol.23 No.4

Asset prices decline sharply and stock markets collapse when financial crisis happens. Recently we have encountered more frequent financial crises than ever. 1998 currency crisis and 2008 global financial crisis triggered academic researches on early warning systems that aim to detect the symptom of financial crisis in advance. This study proposes a risk recovery index for detection of good opportunities from financial market instability. We use SVM classifier algorithms to separate recovery period from unstable financial market data. Input variables are KOSPI index and V-KOSPI200 index. Our SVM algorithms show highly accurate forecasting results on testing data as well as training data. Risk recovery index is derived from our SVM-trained outputs. We develop a trading system that utilizes the suggested risk recovery index. The trading result records very high profit, that is, its annual return runs to 121%.

Seismic Performance of Wind-Designed Diagrid Tall Steel Buildings in Regions of Moderate Seismicity and Strong Wind

김선웅,이경구 국제구조공학회 2013 Steel and Composite Structures, An International J Vol.14 No.2

This study analytically evaluated the seismic performance of wind-designed diagrid tall steel buildings in regions of moderate/low seismicity and strong winds. To this end, diagrid tall steel buildings with varying wind exposure and slenderness ratio (building height-to-width ratio) conditions were designed to satisfy the wind serviceability criteria specified in the Korean Building Code and the National Building Code of Canada. A series of seismic analyses were then performed for earthquakes having 43- and 2475-year return periods utilizing the design guidelines of tall buildings. The analyses demonstrated the good seismic performance of these wind-designed diagrid tall steel buildings, which arises because significant overstrength of the diagrid system occurs in the wind design procedure. Also, analysis showed that the elastic seismic design process of diagrid tall steel buildings might be accepted based on some wind exposures and slenderness ratios.

Potentials of elastic seismic design of twisted high-rise steel diagrid frames

김선웅,이경구 국제구조공학회 2015 Steel and Composite Structures, An International J Vol.18 No.1

This paper is to investigate the potentials of the elastic seismic design of twisted high-rise steel diagrid frame buildings in the strong wind and moderate/low seismicity regions. First, the prototypes of high-rise steel diagrid frames with architectural plans that have a twist angle of 0 (regular-shaped), 1, and 2 degrees were designed to resist wind. Then, the effects of the twist angle on the estimated quantities and structural redundancies of the diagrid frames were examined. Second, the seismic performance of the wind-designed prototype buildings under a low seismicity was evaluated. The response spectrum analysis was conducted for the service level earthquake (SLE) having 43-year return period and the maximum considered earthquake (MCE) having 2475-year return period. The evaluation resulted that the twisted high-rise steel diagrid frames resisted the service level earthquake elastically and most of their diagrid members remained elastic even under the maximum considered earthquake.

Specificity Condition Revisited

김선웅 한국생성문법학회 2013 생성문법연구 Vol.23 No.3

This paper aims to explain the well-known asymmetry of extractability between specific and non-specific NPs. This paper starts with a short review of the previous studies regarding the Specificity Condition from both theoretical and empirical point of view. With the criticism that the condition is nothing but a descriptive generalization which must be deductively derived from deeper principles, this paper explains it in terms of phasehood and its role in the possible domain of extraction. Assuming the dynamic definition of phases (den Dikken 2006, 2007 and Bošković 2012a, 2012b), this paper shows that the ban on extraction from specific NPs is due to violation of the cooperative requirement of the PIC and the Anti-locality. This paper extends the solution to further counterexamples of the Specificity Condition. To be specific, Grimshaw's (1990) notion of event participants are shown to be operative in building additional XPs inside nominals and this addition will change the dynamics of interacting syntactic principles including the PIC and the Anti-locality thereby explaining the counterexamples.

LSTM 기반 COVID-19 공포지수의 주가 예측 성과: 언택트 주식과 콘택트 주식

김선웅 한국콘텐츠학회 2022 한국콘텐츠학회논문지 Vol.22 No.8

As the non-face-to-face economic situation developed due to the COVID-19 pandemic, untact stock groups appeared in the stock market. This study proposed the Korea COVID-19 fear index following the spread of infectious diseases in the COVID-19 pandemic situation and analyzed the influence on the untact stock and contact stock returns. The results of the empirical analysis are as follows. First, as a result of the Granger causality analysis using the Korea COVID-19 fear index, significant causality was found in the return of contact stocks such as Korean Air, Hana Tour, CJ CGV, and Paradise. Second, as a result of stock price prediction based on the LSTM model, Kakao, Korean Air, and Naver's prediction performance was high. Third, the investment performances of the Alexander filter entry rule using the predicted stock price were high in Naver futures and Kakao futures. This study can find a difference from previous studies in that it analyzed the influence of the spread of the COVID-19 pandemic on untact and contact stocks in the COVID-19 situation where the non-face-to-face economy is in full swing. COVID-19 팬데믹으로 비대면 경제 상황이 전개되면서 주식시장에서는 언택트 주식 집단이 등장하였다. 본 연구는 COVID-19 팬데믹 상황에서 감염병 확산에 따른 한국 COVID-19 공포지수를 제안하고, 언택트 주식 수익률과 콘택트 주식 수익률에 대한 영향력을 분석하였다. 실증 분석 결과는 다음과 같다. 첫째, 한국 COVID-19 공포지수를 이용한 그랜저 인과관계 분석 결과 대한항공, 하나투어, CJ CGV, 파라다이스와 같은 콘택트 주식의 수익률에서 유의적인 인과성이 나타났다. 둘째, LSTM 모형 기반의 주가 예측 결과 카카오, 대한항공과 네이버의 예측 성과가 높게 나타났다. 셋째, 예측 주가를 이용한 Alexander 필터 진입 전략의 투자성과는 네이버 선물과 카카오 선물에서 높게 나타났다. 본 연구는 비대면 경제가 본격화된 COVID-19 상황에서 언택트 주식과 콘택트 주식에 대한 COVID-19 팬데믹 확산의 영향력을 분석하였다는 점에서 기존 연구와 차별점을 찾을 수 있다.

특정성과 서술성의 통사론: 영어와 한국어의 (‘것'-)분열문을 중심으로

김선웅 대한언어학회 2010 언어학 Vol.18 No.3

Kim, Sun-Woong. 2010. The Syntax of Specification and Predication: (Kes-)Clefts in Korean and English. The Linguistic Association of Korea Journal. 18(3). 167-185. This paper argues that the Korean kes-cleft construction is a mixture of clefts (it-clefts) and pseudoclefts (wh-clefts) in that it has the properties of both clefts at the same time. To show this, comparisons with English it-clefts and wh-clefts are made with respect to the well-known syntactic and semantic properties. Regarding semantic properties of kes-clefts, it is concluded that the Korean kes-cleft construction is basically of specificational nature rather than predicational. In order to properly derive kes-clefts, following the lead of den Dikken(2009), it is argued that kukes in Korean kes-clefts starts as a small clause predicate and is raised to the front in terms of predicate raising in the sense of Moro(1997). The proposed derivation turns out to effectively explain the binding connectivity and the control possibility of kes-clefts.

Parallel Linearization: A thought on afterthoughts

김선웅,박명관 한국현대언어학회 2010 언어연구 Vol.26 No.3

This paper starts with noting that so-called ‘right dislocation’ in Korean is construed as giving an afterthought, comparable to one of the functions that right dislocation in English takes on (Grosz and Ziv 1996, C.-H. Lee 2009). It is argued below that the right-edge afterthought part of the construction in Korean is derived neither by base-generation (Sells 1999, Yoon 2009), nor by movement (Haraguchi 1973, H.-S. Choe 1987), nor by deletion (Abe 1999), but by the proposed operation of parallel linearization in terms of conjunction reduction (McCawley 1982). In particular, it is shown that the null elements in the afterthought part can be analyzed on a par with copy traces left by movement. To be concrete, the afterthought part of the construction is derived by parallel linearization of the two coordinate clauses: the host and the afterthought clauses. The idea is that as the two clauses undergo linearization in a parallel fashion, the repetitive elements in the second clause undergo phonological reduction. We argue that parallel linearization and reduction are needed anyway for constructions involving conjunction reduction independently of the construction concerned in this paper. We rely on Reinhart's (1976, 1983) Condition on Bound Variable Pronouns to account for NPI elements with respect to afterthoughts.

COVID-19 국면의 암호화폐 가격 예측: 네이버트렌드와 딥러닝의 융합 연구

김선웅 중소기업융합학회 2022 융합정보논문지 Vol.12 No.3

The purpose of this study is to analyze whether investor anxiety caused by COVID-19 affects cryptocurrency prices in the COVID-19 pandemic, and to experiment with cryptocurrency price prediction based on a deep learning model. Investor anxiety is calculated by combining Naver’s Corona search index and Corona confirmed information, analyzing Granger causality with cryptocurrency prices, and predicting cryptocurrency prices using deep learning models. The experimental results are as follows. First, CCI indicators showed significant Granger causality in the returns of Bitcoin, Ethereum, and Lightcoin. Second, LSTM with CCI as an input variable showed high predictive performance. Third, Bitcoin's price prediction performance was the highest in comparison between cryptocurrencies. This study is of academic significance in that it is the first attempt to analyze the relationship between Naver’s Corona search information and cryptocurrency prices in the Corona phase. In future studies, extended studies into various deep learning models are needed to increase price prediction accuracy. 본 연구의 목적은 COVID-19 팬데믹 국면에서 코로나 발생과 확산에 따른 투자자 불안심리가 암호화폐 가격에 영향을 미치는지를 분석하고, 딥러닝 모형에 기반하여 암호화폐의 가격 예측을 실험하는 것이다. 투자자 불안심리는 네이버의 코로나 검색지수와 코로나 확진자 정보를 결합하여 산출하며, 암호화폐 가격과의 그랜저 인과성을 분석하고 딥러닝모형을 이용하여 암호화폐 가격을 예측한다. 실험 결과는 다음과 같다. 첫째, CCI 지표는 비트코인, 이더리움, 라이트코인의 수익률에 유의적인 그랜저 인과성을 보여주었다. 둘째, CCI를 입력변수로 하는 LSTM은 높은 예측성과를 보여주었다. 셋째, 암호화폐 사이의 비교에서는 비트코인의 가격 예측 성과가 가장 높게 나타났다. 본 연구는 코로나 국면에서 네이버 코로나 검색 정보와 암호화폐 가격과의 관련성을 분석한 첫 시도라는 점에서 학술적 의의를 찾을 수 있다. 향후 연구에서는 가격 예측 정확성을 높이기 위하여 다양한 딥러닝 모형으로의 확장 연구가 필요하다.

Passivation of Semi-Insulating Polycrystaline CdZnTe Films

김선웅 한국물리학회 2008 THE JOURNAL OF THE KOREAN PHYSICAL SOCIETY Vol.53 No.1

SurfaceeectsplayanimportantroleintheoveralperformanceofX-raydetector. Theeectsof pasivationwith(NH4)2Sonsemi-insulatingpolycrystalineCdZnTethicklmswereanalyzedwith X-rayphotoelectronspectroscopy(XPS),photoconductivedecay(PCD),noisepowerspectrumand pulse height spectra measurements. Sulfur pasivation with (HN4)2S eectively removes the Te- oxide layers on the CdZnTe surface, reduces the surface leakage curent and gives higher energy resolutionbysuppresing1=f noise.

On the Relationship between Linear Proximity and Subject Verb Agreement Mismatch in Korean L2 Learners of English: An ERP Study

김선웅,박명관,정원일 한국생성문법학회 2012 생성문법연구 Vol.22 No.3

This paper claims that not only linear proximity but also other various factors including structural and semantic composition of the subject NP can affect the subject-verb agreement (SVA) mismatch in L2 learning of English. Based on the previous researches on both experimental and theoretical studies on language, this paper proposes that what is genuinely responsible for the SVA mismatch due to proximity is actually a reflection of the multiple activation of potential subject NPs. This multiple activation is claimed to be the source of confusion, of which the notion has been found to be true on the side of previous experimental studies. This paper conducted two ERP experiments to see whether proficient Korean L2 learners of English also experience the same confusion in the case of SVA mismatch as L1. One is to see whether a more proximate NP also exerts an influence on agreement in L2 setting as has been known to be the case in L1. The other is to see whether L2 speakers take into consideration for agreement the semantic or pragmatic closeness between the head noun and the relevant potential NP. Through Experiment 1, it was found that proficient Korean L2 learners of English showed a LAN as well as a late P600 effect, which evidences that L2 learners are also experiencing the confusion due to the multiple activation of potential subject NPs. Experiment 2, however, gave no clue of outright semantic influence on the SVA mismatch, but a marginal P600 was found when they met a false form of the verb be. This means that semantic/pragmatic factors do not exert an influence on determining the agreement verb form, but some process of syntactic reanalysis may be in operation in L2 learning.