• Radiation Oncology Journal
• /
• 제32권3호
• /
• pp.179-186
• /
• 2014

Purpose: To study the long-term outcomes and tolerance in our patients who received dose escalated radiotherapy in the early salvage post-prostatectomy setting. Materials and Methods: The medical records of 54 consecutive patients who underwent radical prostatectomy subsequently followed by salvage radiation therapy (SRT) to the prostate bed between 2003-2010 were analyzed. Patients included were required to have a pre-radiation prostate specific antigen level (PSA) of 2 ng/mL or less. The median SRT dose was 70.2 Gy. Biochemical failure after salvage radiation was defined as a PSA level >0.2 ng/mL. Biochemical control and survival endpoints were analyzed using the Kaplan-Meier method. Univariate and multivariate Cox regression analysis were used to identify the potential impact of confounding factors on outcomes. Results: The median pre-SRT PSA was 0.45 ng/mL and the median follow-up time was 71 months. The 4- and 7-year actuarial biochemical control rates were 75.7% and 63.2%, respectively. The actuarial 4- and 7-year distant metastasis-free survival was 93.7% and 87.0%, respectively, and the actuarial 7-year prostate cancer specific survival was 94.9%. Grade 3 late genitourinary toxicity developed in 14 patients (25.9%), while grade 4 late genitourinary toxicity developed in 2 patients (3.7%). Grade 3 late gastrointestinal toxicity developed in 1 patient (1.9%), and grade 4 late gastrointestinal toxicity developed in 1 patient (1.9%). Conclusion: In this series with long-term follow-up, early SRT provided outcomes and toxicity profiles similar to those reported from the three major randomized trials studying adjuvant radiation therapy.

• Radiation Oncology Journal
• /
• 제31권4호
• /
• pp.234-238
• /
• 2013

Purpose: Esophageal tolerance is needed to guide the safe administration of stereotactic radiosurgery (SRS). We evaluated comprehensive dose-volume parameters of acute esophageal toxicity in patients with spinal metastasis treated with SRS. Materials and Methods: From May 2008 to May 2011, 30 cases in 27 patients with spinal metastasis received single fraction SRS to targets neighboring esophagus. Endpoints evaluated include length (mm), volume (mL), maximal dose (Gy), and series of dose-volume thresholds from the dose-volume histogram (volume of the organ treated beyond a threshold dose). Results: The median time from the start of irradiation to development of esophageal toxicity was 2 weeks (range, 1 to 12 weeks). Six events of grade 1 esophageal toxicity occurred. No grade 2 or higher events were observed. $V_{15}$ of external surface of esophagus was found to predict acute esophageal toxicity revealed by multivariate analysis (odds radio = 1.272, p = 0.047). Conclusion: In patients with spinal metastasis who received SRS for palliation of symptoms, the threshold dose-volume parameter associated with acute esophageal toxicity was found to be $V_{15}$ of external surface of esophagus. Restrict $V_{15}$ to external surface of esophagus as low as possible might be safe and feasible in radiosurgery.

• 통상정보연구
• /
• 제16권3호
• /
• pp.73-96
• /
• 2014

본 연구는 우리나라 수출 상위 5개 품목 중 하나인 자동차 수출을 대상으로, 승용차 브랜드별 단기 수출수요에 영향을 미치는 이론적 잠재요인을 발굴 및 설계하여 이론적 수출수요예측모델을 개발하고, 다변량시계열분석 기반의 VAR(Vector Auto Regressive)모형을 이용한 실증분석을 통해 개별상품과 시장특성이 반영된 단기수출수요예측모델을 검정하고자 하였다. 따라서 미국에 수출되고 있는 우리나라 소형 승용차 2개 브랜드(엑센트, 아반떼)에 대해 VAR모형을 이용한 분기단위 단기수요예측모델을 개발하고, 브랜드별 예측모델을 통해 산출된 t+1분기 시점의 예측값과 실제 판매된 판매대수를 대상기간을 1분기씩 달리하여 비교평가 하였다. 그 결과 엑센트와 아반떼의 RMSE %는 각각 4.3%와 20.0%로 났으며, 일평균 판매량을 기준으로 보았을 때 엑센트는 3.9일에 해당하고 아반떼는 18.4일에 해당하는 물량임을 알 수 있었다. 따라서 본 연구의 단기수출수요예측모델은 예측력과 검정시점별 일관성 측면에서 활용성이 높은 것으로 평가할 수 있었다.

• Journal of the Korean Data and Information Science Society
• /
• 제25권6호
• /
• pp.1449-1466
• /
• 2014

본 논문에서는 기초자산의 선물을 이용하는 헷지 전략을 연구하였다. 최적헷지비율을 구하기 위한 전통적인 방법으로 회귀분석이 사용되고 있으나, 현물과 선물 사이에 존재하는 장기균형관계와 금융 시계열 자료의 분산에 존재하는 변동성 군집현상 등의 특징을 설명하지 못하는 한계가 있다. 이를 극복하기 위해 코스피200 지수와 선물 자료에 대해 평균모형으로 벡터오차수정모형을 적합하고, 분산모형으로 다변량 GARCH 모형을 적합하여 분산-공분산 행렬을 추정하고, 이를 통해 최적헷지비율을 구하는 방법을 연구하였다. 실증분석 결과에 의하면 시장이 안정적일 때에는 회귀분석을 사용해도 큰 차이가 없지만, 시장이 불안정해지고 변동성이 커지는 구간에서는 벡터오차수정모형과 다변량 GARCH 모형을 이용하는 경우에 헷지성과가 월등히 좋아지는 결과를 얻을 수 있었다.

• International Journal of Aeronautical and Space Sciences
• /
• 제13권1호
• /
• pp.14-26
• /
• 2012

The only extraterrestrial life detection experiments ever conducted were the three which were components of the 1976 Viking Mission to Mars. Of these, only the Labeled Release experiment obtained a clearly positive response. In this experiment $^{14}C$ radiolabeled nutrient was added to the Mars soil samples. Active soils exhibited rapid, substantial gas release. The gas was probably $CO_2$ and, possibly, other radiocarbon-containing gases. We have applied complexity analysis to the Viking LR data. Measures of mathematical complexity permit deep analysis of data structure along continua including signal vs. noise, entropy vs.negentropy, periodicity vs. aperiodicity, order vs. disorder etc. We have employed seven complexity variables, all derived from LR data, to show that Viking LR active responses can be distinguished from controls via cluster analysis and other multivariate techniques. Furthermore, Martian LR active response data cluster with known biological time series while the control data cluster with purely physical measures. We conclude that the complexity pattern seen in active experiments strongly suggests biology while the different pattern in the control responses is more likely to be non-biological. Control responses that exhibit relatively low initial order rapidly devolve into near-random noise, while the active experiments exhibit higher initial order which decays only slowly. This suggests a robust biological response. These analyses support the interpretation that the Viking LR experiment did detect extant microbial life on Mars.

• 품질경영학회지
• /
• 제49권4호
• /
• pp.581-594
• /
• 2021

Purpose: The purpose of this study is to compare machine learning models for anomaly detection of mechanical facility equipment and suggest an anomaly detection system for mechanical facility equipment in subway stations. It helps to predict failures and plan the maintenance of facility. Ultimately it aims to improve the quality of facility equipment. Methods: The data collected from Daejeon Metropolitan Rapid Transit Corporation was used in this experiment. The experiment was performed using Python, Scikit-learn, tensorflow 2.0 for preprocessing and machine learning. Also it was conducted in two failure states of the equipment. We compared and analyzed five unsupervised machine learning models focused on model Long Short-Term Memory Variational Autoencoder(LSTM-VAE). Results: In both experiments, change in vibration and current data was observed when there is a defect. When the rotating body failure was happened, the magnitude of vibration has increased but current has decreased. In situation of axis alignment failure, both of vibration and current have increased. In addition, model LSTM-VAE showed superior accuracy than the other four base-line models. Conclusion: According to the results, model LSTM-VAE showed outstanding performance with more than 97% of accuracy in the experiments. Thus, the quality of mechanical facility equipment will be improved if the proposed anomaly detection system is established with this model used.

• Smart Structures and Systems
• /
• 제29권1호
• /
• pp.117-127
• /
• 2022

Data anomalies seriously threaten the reliability of the bridge structural health monitoring system and may trigger system misjudgment. To overcome the above problem, an efficient and accurate data anomaly detection method is desiderated. Traditional anomaly detection methods extract various abnormal features as the key indicators to identify data anomalies. Then set thresholds artificially for various features to identify specific anomalies, which is the artificial experience method. However, limited by the poor generalization ability among sensors, this method often leads to high labor costs. Another approach to anomaly detection is a data-driven approach based on machine learning methods. Among these, the bidirectional long-short memory neural network (BiLSTM), as an effective classification method, excels at finding complex relationships in multivariate time series data. However, training unprocessed original signals often leads to low computation efficiency and poor convergence, for lacking appropriate feature selection. Therefore, this article combines the advantages of the two methods by proposing a deep learning method with manual experience statistical features fed into it. Experimental comparative studies illustrate that the BiLSTM model with appropriate feature input has an accuracy rate of over 87-94%. Meanwhile, this paper provides basic principles of data cleaning and discusses the typical features of various anomalies. Furthermore, the optimization strategies of the feature space selection based on artificial experience are also highlighted.

• Nuclear Engineering and Technology
• /
• 제55권9호
• /
• pp.3423-3440
• /
• 2023

In this work, a multivariate time-series machine learning meta-model is developed to predict the transient response of a typical nuclear power plant (NPP) undergoing a steam generator tube rupture (SGTR). The model employs Recurrent Neural Networks (RNNs), including the Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and a hybrid CNN-LSTM model. To address the uncertainty inherent in such predictions, a Bayesian Neural Network (BNN) was implemented. The models were trained using a database generated by the Best Estimate Plus Uncertainty (BEPU) methodology; coupling the thermal hydraulics code, RELAP5/SCDAP/MOD3.4 to the statistical tool, DAKOTA, to predict the variation in system response under various operational and phenomenological uncertainties. The RNN models successfully captures the underlying characteristics of the data with reasonable accuracy, and the BNN-LSTM approach offers an additional layer of insight into the level of uncertainty associated with the predictions. The results demonstrate that LSTM outperforms GRU, while the hybrid CNN-LSTM model is computationally the most efficient. This study aims to gain a better understanding of the capabilities and limitations of machine learning models in the context of nuclear safety. By expanding the application of ML models to more severe accident scenarios, where operators are under extreme stress and prone to errors, ML models can provide valuable support and act as expert systems to assist in decision-making while minimizing the chances of human error.

• 지질공학
• /
• 제23권4호
• /
• pp.353-361
• /
• 2013

본 연구에서는 ${\bigcirc}{\bigcirc}$지역 토목용 계측기에서 측정된 지진유발 변형률 데이터의 분포 특성을 분석하기 위한 기법으로 응용통계기법에 대한 적용성을 평가하였다. 2011년 도호쿠 대지진과 같은 해에 발생한 규모 7.0 이상의 여진을 계측한 4방향의 변형률 데이터를 활용하였다. 데이터의 미세한 변동을 감지하기 위하여 단변량 분석기법인 x-MR 분석을 실시하였으며 분석결과 계측 데이터 간의 분산시점에 차이가 발생하는 것을 확인하였다. 이러한 분산시점의 차이를 해결하기 위하여 변형률 데이터 간의 상관성을 고려한 다변량 통계분석을 실시하였다. 다변량 분석기법 가운데 하나인 주성분 분석결과를 $T_2$과 Q-통계량 분석에 적용하여 신뢰구간 99.9%, 99.0%, 95.0%로 실시간 분석을 수행하였다. 분석결과 $T_2$과 Q-통계량 값이 신뢰구간 99.9%를 초과하는 시점은 x-MR 분석의 분산시점과 일치하거나 이른 시간으로 나타났다. 또한, 신뢰구간 95.0%와 99.0%를 초과하는 시점은 99.9%를 초과하는 시점 이전에 타점되어 지진발생 전에 이상 분포 발생을 예측할 수 있었다. 이러한 결과는 변형률 데이터의 비정상적인 분포 특성을 다변량 통계분석법으로 인지할 수 있다는 것을 의미한다. 따라서 다변량 통계분석은 변형률 데이터의 분포 특성을 분석하여 지진을 예지하는 방법으로 이용가능하다고 판단된다.

• 한국지구과학회지
• /
• 제29권1호
• /
• pp.29-44
• /
• 2008

본 연구에서는 El $Ni{\tilde{n}}o$-Southern Oscillation(ENSO) 발달과 소멸의 영향에 따른 태풍 강도와 태풍 발생지역의 상관성을 살펴보았다. 1950년부터 2006년까지의 장기간 자료를 이용하였으며, 먼저 엘니뇨 발달해와 정상해를 정의하였다. 엘니뇨 발달해 동안에 태풍 강도와 태풍 발생지역이 높은 상관성을 나타내고 이는 누적 저기압 에너지와 태풍 에너지 강도가 증가한 결과이다. $Ni{\tilde{n}}o$ 3.4 지역의 해수면 온도를 기준으로 한 경우 엘니뇨 발달해에는, category 4+5에 해당하는 태풍의 발생지역이 동쪽으로 치우쳐 나타난다. 태풍 발생 잠재 함수와 하층의 저기압성 회전성은 태풍급에 해당하는 강도로 발달할 수 있는 강한 열대성 저기압의 발생에 중요한 요소가 된다. 본 논문에서는 역학적 잠재력[DP, Gray(1977)]과 MJO의 EOF 첫 번째 모드와 두 번째 모드의 시계열에 해당하는 RMM1, RMM2 (Wheeler and Hendon, 2004)를 이용하여 태풍 발생의 잠재함수와 대기 하층의 저기압성 회전성을 측정하였다. ENSO가 발달하는 해와 소멸하는 해와 영향을 찾아보기 위하여 엘니뇨가 소멸이 급격히 일어나 라니냐로 전환되는 Type I과 정상해로 회복하는 Type II를 정의하였다. Type I의 엘니뇨 소멸기간 동안에는 DP값과 RMM1, RMM2의 발달이 현저하게 서쪽으로 치우쳐 나타나며 강한 태풍의 발달을 지체시킴을 알 수 있었다.