• 한국수자원학회논문집
• /
• 제54권spc1호
• /
• pp.1183-1193
• /
• 2021

Low flow는 하천수의 공급관리 및 계획, 관개용수 등 다양한 분야에 영향을 미친다. 이러한 유황곡선을 산정하기 위해서는 30년 이상의 충분한 기간의 유량자료의 확보가 필수적이다. 하지만 국가하천 단위 이하의 하천의 경우 장기간의 유량자료가 없거나 중간에 일정기간 동안 결측된 관측소가 있어 하천별 유황 곡선을 산정하기에 한계가 있다. 이에 과거에는 미계측 유역의 유황을 예측하기 위해 다중회귀분석(Multiple Regression Analysis), ARIMA 모형 등 통계학적 기반의 기법들을 사용하였지만, 최근에는 머신러닝, 딥러닝 모형의 수요가 증가하고 있다. 이에 본 연구에서는 최신 패러다임에 맞는 머신러닝 기법인 DNN기법을 제시한다. DNN기법은 ANN기법의 단점인 학습과정에서 최적 매개변수 값을 찾기 어렵고, 학습시간이 느린 단점을 보완한 방법이다. 따라서 본연구에서는 DNN 모형을 이용하여 미계측 유역에 적용 가능한 유황곡선을 산정하고자 한다. 먼저, 유황곡선에 영향을 미치는 인자들을 수집하고 인자들 간의 다중공선성 분석을 통해 통계적으로 유의한 변수를 선정하여, 머신러닝 모형에 입력자료를 구축하였다. 통계적 검증을 통해 머신러닝 기법의 효용성을 검토하였다.

• 한국빅데이터학회지
• /
• 제4권2호
• /
• pp.61-71
• /
• 2019

반도체 웨이퍼의 표면을 연마하여 평탄화하는 Chemical Mechanical Planarization(CMP) 공정은 다양한 화학물질과 물리적인 기계장치에 의한 작용을 받기 때문에 공정을 안정적으로 관리하기 힘들다. CMP 공정에서 품질 지표로는 Material Removal Rate(MRR)를 많이 사용하고, CMP 공정의 안정적 관리를 위해서는 MRR을 예측하는 것이 중요하다. 본 연구에서는 머신러닝 기법들을 이용하여 CMP 공정에서 수집된 시계열 센서 데이터를 분석하여 MRR을 예측하는 모형과 공정 품질을 해석하기 위한 분류 모형을 개발한다. 나아가 분류 결과를 분석하여, CMP 공정 품질에 영향을 미치는 유의미한 변수를 파악하고 고품질을 유지하기 위한 공정 조건을 설명한다.

• 반도체디스플레이기술학회지
• /
• 제19권1호
• /
• pp.36-41
• /
• 2020

In this paper, We need to change all print factors when which print scaffold with 400 ㎛ pore using FDM 3d printer. Therefore the print quantity is 10 billion times, So we are difficult to print on workplace. To solve the problem, we used the prediction model based machine learning regression. We preprocessed and learned the securing print condition data, and we produced different kinds of prediction models. We predicted the pore size of scaffolds not securing with new print condition data using prediction models. We have derived the print conditions that satisfy the pore size of 400 ㎛ among the predicted print conditions of pore size. We printed the scaffolds 5 times on the condition. We measured the pore size of the printed scaffold and compared the average pore size with the predicted pore size. We confirmed that error was less than 1%, and we were identify the model with the highest pore size prediction performance of scaffold.

• 한국해양공학회지
• /
• 제35권5호
• /
• pp.336-346
• /
• 2021

The prediction of wave conditions is crucial in the field of marine and ocean engineering. Hence, this study aims to predict the significant wave height through machine learning (ML), a soft computing method. The adopted metocean data, collected from 2012 to 2020, were obtained from the Korea Institute of Ocean Science and Technology. We adopted the feedforward neural network (FNN) and long-short term memory (LSTM) models to predict significant wave height. Input parameters for the input layer were selected by Pearson correlation coefficients. To obtain the optimized hyperparameter, we conducted a sensitivity study on the window size, node, layer, and activation function. Finally, the significant wave height was predicted using the FNN and LSTM models, by varying the three input parameters and three window sizes. Accordingly, FNN (W48) (i.e., FNN with window size 48) and LSTM (W48) (i.e., LSTM with window size 48) were superior outcomes. The most suitable model for predicting the significant wave height was FNN(W48) owing to its accuracy and calculation time. If the metocean data were further accumulated, the accuracy of the ML model would have improved, and it will be beneficial to predict added resistance by waves when conducting a sea trial test.

• KEPCO Journal on Electric Power and Energy
• /
• 제7권1호
• /
• pp.171-177
• /
• 2021

In terms of distribution planning, accurate electric load prediction is one of the most important factors. The future load prediction has manually been performed by calculating the maximum electric load considering loads transfer/switching and multiplying it with the load increase rate. In here, the risk of human error is inherent and thus an automated maximum electric load forecasting system is required. Although there are many existing methods and techniques to predict future electric loads, such as regression analysis, many of them have limitations in reflecting the nonlinear characteristics of the electric load and the complexity due to Photovoltaics (PVs), Electric Vehicles (EVs), and etc. This study, therefore, proposes a method of predicting future electric loads on distribution lines by using Machine Learning (ML) method that can reflect the characteristics of these nonlinearities. In addition, predictive models were developed based on actual data collected at KEPCO's existing distribution lines and the adequacy of developed models was verified as well. Also, as the distribution planning has a direct bearing on the investment, and amount of investment has a direct bearing on the maximum electric load, various baseline such as maximum, lowest, median value that can assesses the adequacy and accuracy of proposed ML based electric load prediction methods were suggested.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권12호
• /
• pp.4476-4491
• /
• 2021

Video service providers tend to face user network problems in the process of transmitting video streams. They strive to provide user with superior video quality in a limited bitrate environment. It is necessary to accurately determine the target bitrate range of the video under different quality requirements. Recently, several schemes have been proposed to meet this requirement. However, they do not take the impact of visual influence into account. In this paper, we propose a new multi-category model to accurately predict the target bitrate range with target visual quality by machine learning. Firstly, a dataset is constructed to generate multi-category models by machine learning. The quality score ladders and the corresponding bitrate-interval categories are defined in the dataset. Secondly, several types of spatial-temporal features related to VMAF evaluation metrics and visual factors are extracted and processed statistically for classification. Finally, bitrate prediction models trained on the dataset by RandomForest classifier can be used to accurately predict the target bitrate of the input videos with target video quality. The classification prediction accuracy of the model reaches 0.705 and the encoded video which is compressed by the bitrate predicted by the model can achieve the target perceptual quality.

• Steel and Composite Structures
• /
• 제48권5호
• /
• pp.531-545
• /
• 2023

Composite dowels are implemented as a powerful alternative to headed studs for the efficient combination of Ultra High-Performance Concrete (UHPC) with high-strength steel in novel composite structures. They are required to provide sufficient shear resistance and ensure the transmission of tensile forces in the composite connection in order to prevent lifting of the concrete slab. In this paper, the load bearing capacity of puzzle-shaped and clothoidal-shaped dowels encased in UHPC specimen were investigated based on validated experimental test data. Considering the influence of the embedment depth and the spacing width of shear dowels, the characteristics of UHPC square plate on the load bearing capacity of composite structure, 240 numeric models have been constructed and analyzed. Three artificial intelligence approaches have been implemented to learn the discipline from collected experimental data and then make prediction, which includes Artificial Neural Network-Particle Swarm Optimization (ANN-PSO), Adaptive Neuro-Fuzzy Inference System (ANFIS) and an Extreme Learning Machine (ELM). Among the factors, the embedment depth of composite dowel is proved to be the most influential parameter on the load bearing capacity. Furthermore, the results of the prediction models reveal that ELM is capable to achieve more accurate prediction.

• Computers and Concrete
• /
• 제31권4호
• /
• pp.277-292
• /
• 2023

Much of our experiments are designed to uncover the cause(s) and effect(s) behind a phenomenon (i.e., data generating mechanism) we happen to be interested in. Uncovering such relationships allows us to identify the true workings of a phenomenon and, most importantly, to realize and articulate a model to explore the phenomenon on hand and/or allow us to predict it accurately. Fundamentally, such models are likely to be derived via a causal approach (as opposed to an observational or empirical mean). In this approach, causal discovery is required to create a causal model, which can then be applied to infer the influence of interventions, and answer any hypothetical questions (i.e., in the form of What ifs? Etc.) that commonly used prediction- and statistical-based models may not be able to address. From this lens, this paper builds a case for causal discovery and causal inference and contrasts that against common machine learning approaches - all from a civil and structural engineering perspective. More specifically, this paper outlines the key principles of causality and the most commonly used algorithms and packages for causal discovery and causal inference. Finally, this paper also presents a series of examples and case studies of how causal concepts can be adopted for our domain.

• 생물정신의학
• /
• 제30권1호
• /
• pp.24-30
• /
• 2023

Objectives Alzheimer's disease (AD) is the most common form of dementia in older adults, damaging the brain and resulting in impaired memory, thinking, and behavior. The identification of differentially expressed genes and related pathways among affected brain regions can provide more information on the mechanisms of AD. The aim of our study was to identify differentially expressed genes associated with AD and combined biomarkers among them to improve AD risk prediction accuracy. Methods Machine learning methods were used to compare the performance of the identified combined biomarkers. In this study, three publicly available gene expression datasets from the hippocampal brain region were used. Results We detected 31 significant common genes from two different microarray datasets using the limma package. Some of them belonged to 11 biological pathways. Combined biomarkers were identified in two microarray datasets and were evaluated in a different dataset. The performance of the predictive models using the combined biomarkers was superior to those of models using a single gene. When two genes were combined, the most predictive gene set in the evaluation dataset was ATR and PRKCB when linear discriminant analysis was applied. Conclusions Combined biomarkers showed good performance in predicting the risk of AD. The constructed predictive nomogram using combined biomarkers could easily be used by clinicians to identify high-risk individuals so that more efficient trials could be designed to reduce the incidence of AD.

• 한국분말재료학회지
• /
• 제30권3호
• /
• pp.210-216
• /
• 2023

In this study, machine learning models are proposed to predict the Vickers hardness of AlSi10Mg alloys fabricated by laser powder bed fusion (LPBF). A total of 113 utilizable datasets were collected from the literature. The hyperparameters of the machine-learning models were adjusted to select an accurate predictive model. The random forest regression (RFR) model showed the best performance compared to support vector regression, artificial neural networks, and k-nearest neighbors. The variable importance and prediction mechanisms of the RFR were discussed by Shapley additive explanation (SHAP). Aging time had the greatest influence on the Vickers hardness, followed by solution time, solution temperature, layer thickness, scan speed, power, aging temperature, average particle size, and hatching distance. Detailed prediction mechanisms for RFR are analyzed using SHAP dependence plots.