• Journal of the Korean earth science society
• /
• v.23 no.7
• /
• pp.587-596
• /
• 2002

A super ensemble model was developed for the seasonal prediction of regional precipitation in Korea using the lag correlated large scale predictors, based on the empirical orthogonal function (EOF) analysis and multiple linear regression model. The predictability of this model was also evaluated by cross-validation. Correlation between the predicted and the observed value obtained from the super ensemble model showed 0.73 in spring, 0.61 in summer, 0.69 in autumn and 0.75 in winter. The predictability of categorical forecasting was also evaluated based on the three classes such as above normal, near normal and below normal that are clearly defined in terms of a priori specified by threshold values. Categorical forecasting by the super ensemble model has a hit rate with a range from 0.42 to 0.74 in seasonal precipitation.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2017.05a
• /
• pp.135-135
• /
• 2017

최근 강우-유출 모형은 물리적 현상에 근거한 확정론적 모의 모형과 물리적 성분으로 설명할 수 없는 내용에 대해 통계적으로 접근하는 추계학적 모의 모형 등이 계속 연구되고 있어 자연현상에 가까운 결과를 기대할 수 있게 되었다. 하지만 우리나라의 경우 많은 연구에도 불구하고 돌발성 집중호우, 여름철 집중되는 강우 등으로 인해 재난이 반복적으로 발생하고 있어 모형의 정확성에 대한 논의가 지속되고 있다. 동일한 유역에 동일한 입력자료를 사용하더라도 사용하는 모형에 따라 유출 분석결과는 상이하며 이는 유출 해석에 대한 불확실성으로 작용한다. 본 연구에서는 앙상블 및 블랜딩 기법을 사용하여 각 강우-유출 모형의 불확실성을 고려하여 최적 유출량을 산정하고자 한다. 대상 유역으로는 한강 수계에 있는 중랑천 유역을 선정하였으며, Distributed 모형인 Vflo 모형과 Lumped 모형인 저류함수 모형, SSARR모형, TANK 모형을 이용하여 유출 분석을 실시하였다. 그 후, Multi-Model Super Ensemble(MMSE), Simple Model Average(SMA), Mean Square Error(MSE) 방법 등의 blending 기법을 이용하여 하나의 통합된 형태의 유출 분석 결과를 제시하였으며, 최적 유출량 산정을 위한 blending 기법을 선정하였다. 본 연구를 통해 동일한 강우 시나리오에 대한 여러 강우-유출 모형에 대한 정확도를 확인하였으며, 앙상블 및 블랜딩 기법을 사용하여 유출 분석에 대한 정확도를 향상시킬 수 있을 것으로 판단된다.

• Geophysics and Geophysical Exploration
• /
• v.25 no.3
• /
• pp.140-161
• /
• 2022

Seismic data with missing traces are often obtained regularly or irregularly due to environmental and economic constraints in their acquisition. Accordingly, seismic data interpolation is an essential step in seismic data processing. Recently, research activity on machine learning-based seismic data interpolation has been flourishing. In particular, convolutional neural network (CNN) and generative adversarial network (GAN), which are widely used algorithms for super-resolution problem solving in the image processing field, are also used for seismic data interpolation. In this study, CNN-based algorithm, U-Net and GAN-based algorithm, and conditional Wasserstein GAN (cWGAN) were used as seismic data interpolation methods. The results and performances of the methods were evaluated thoroughly to find an optimal interpolation method, which reconstructs with high accuracy missing seismic data. The work process for model training and performance evaluation was divided into two cases (i.e., Cases I and II). In Case I, we trained the model using only the regularly sampled data with 50% missing traces. We evaluated the model performance by applying the trained model to a total of six different test datasets, which consisted of a combination of regular, irregular, and sampling ratios. In Case II, six different models were generated using the training datasets sampled in the same way as the six test datasets. The models were applied to the same test datasets used in Case I to compare the results. We found that cWGAN showed better prediction performance than U-Net with higher PSNR and SSIM. However, cWGAN generated additional noise to the prediction results; thus, an ensemble technique was performed to remove the noise and improve the accuracy. The cWGAN ensemble model removed successfully the noise and showed improved PSNR and SSIM compared with existing individual models.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.22 no.6
• /
• pp.36-42
• /
• 2021

With recent developments in the Fourth Industrial Revolution, the manufacturing industry has changed rapidly. Through key aspects of Fourth Industrial Revolution super-connections and super-intelligence, machine learning will be able to make fault predictions during the foam-making process. Polyol and isocyanate are components in polyurethane foam. There has been a lot of research that could affect the characteristics of the products, depending on the specific mixture ratio and temperature. Based on these characteristics, this study collects data from each factor during the foam-making process and applies them to machine learning in order to predict faults. The algorithms used in machine learning are the decision tree, kNN, and an ensemble algorithm, and these algorithms learn from 5,147 cases. Based on 1,000 pieces of data for validation, the learning results show up to 98.5% accuracy using the ensemble algorithm. Therefore, the results confirm the faults of currently produced parts by collecting real-time data from each factor during the foam-making process. Furthermore, control of each of the factors may improve the fault rate.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2021.06a
• /
• pp.140-140
• /
• 2021

Urban flood management is a crucial and challenging task, particularly in developed cities. Therefore, accurate prediction of urban flooding under heavy precipitation is critically important to address such a challenge. In recent years, machine learning techniques have received considerable attention for their strong learning ability and suitability for modeling complex and nonlinear hydrological processes. Moreover, a survey of the published literature finds that hybrid computational intelligent methods using nature-inspired algorithms have been increasingly employed to predict or simulate the streamflow with high reliability. The present study is aimed to propose a novel approach, an ensemble tree, Bayesian Additive Regression Trees (BART) model incorporating a nature-inspired algorithm to predict hourly multi-step ahead streamflow. For this reason, a hybrid intelligent model was developed, namely GA-BART, containing BART model integrating with Genetic algorithm (GA). The Jungrang urban basin located in Seoul, South Korea, was selected as a case study for the purpose. A database was established based on 39 heavy rainfall events during 2003 and 2020 that collected from the rain gauges and monitoring stations system in the basin. For the goal of this study, the different step ahead models will be developed based in the methods, including 1-hour, 2-hour, 3-hour, 4-hour, 5-hour, and 6-hour step ahead streamflow predictions. In addition, the comparison of the hybrid BART model with a baseline model such as super vector regression models is examined in this study. It is expected that the hybrid BART model has a robust performance and can be an optional choice in streamflow forecasting for urban basins.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.37 no.1
• /
• pp.9-17
• /
• 2017

Characteristics of aerodynamic damping ratios of a helical $180^{\circ}$ model which shows better aerodynamic behavior in both along-wind and across-wind responses on a super tall building was investigated by an aeroelastic model test. The aerodynamic damping ratio was evaluated from the wind-induced responses of the model by using Random Decrement (RD) technique. Further, various triggering levels in evaluation of aerodynamic damping ratios using RD technique were also examined. As a result, it was found that when at least 2000 segments were used for evaluating aerodynamic damping ratio for ensemble averaging, the aerodynamic damping ratio can be obtained more consistently with lower irregular fluctuations. This is good agreement with those of previous studies. Another notable observation was that for square and helical $180^{\circ}$ models, the aerodynamic damping ratios in along-wind direction showed similar linear trends with reduced wind speeds regarding of building shapes. On the other hand, for the helical $180^{\circ}$ model, the aerodynamic damping ratio in across-wind direction showed quite different trends with those of the square model. In addition, the aerodynamic damping ratios of the helical $180^{\circ}$ model showed very similar trends with respect to the change of wind direction, and showed gradually increasing trends having small fluctuations with reduced wind speeds. Another observation was that in definition of triggering levels in RD technique on aerodynamic damping ratios, it may be possible to adopt the triggering levels of "standard deviation" or "${\sqrt{2}}$ times of the standard deviation" of the response time history if RD functions have a large number of triggering points. Further, these triggering levels may result in similar values and distributions with reduced wind speeds and either may be acceptable.

• The Korean Journal of Applied Statistics
• /
• v.34 no.3
• /
• pp.309-327
• /
• 2021

Dynamic treatment regimes (DTRs) are decision-making rules designed to provide personalized treatment to individuals in multi-stage randomized trials. Unlike classical methods, in which all individuals are prescribed the same type of treatment, DTRs prescribe patient-tailored treatments which take into account individual characteristics that may change over time. The Q-learning method, one of regression-based algorithms to figure out optimal treatment rules, becomes more popular as it can be easily implemented. However, the performance of the Q-learning algorithm heavily relies on the correct specification of the Q-function for response, especially in observational studies. In this article, we examine a number of double-robust weighted least-squares estimating methods for Q-learning in high-dimensional settings, where treatment models for propensity score and penalization for sparse estimation are also investigated. We further consider flexible ensemble machine learning methods for the treatment model to achieve double-robustness, so that optimal decision rule can be correctly estimated as long as at least one of the outcome model or treatment model is correct. Extensive simulation studies show that the proposed methods work well with practical sample sizes. The practical utility of the proposed methods is proven with real data example.