In this paper, researchers investigated current fashion brands’ social media micro-blogging marketing strategy and consumers’ word of mouth reactions. More than 5,000 of the micro-blogs posted by fashion brands and 143,000 of customers’ comments were analysed in this study. Researchers firstly investigated the overall micro-blogging marketing structure for each fashion brand and compared them. Then researchers identified the type of expression pattern of each fashion brand currently in the micro-blogging context, negative or positive sentiments. Researchers found that fashion brands are using different micro-blogging marketing strategies. Forever21 used 61% micro-blogs for customer communication. H&M posted very diverse micro-blogs content to their official account. Their main micro-blogs were used for new product promotion (43%) and brand’s live event broadcasting (33%). Luxury brands, such as Burberry, more than 52% of micro-blogs posted last year were used for new product promotion and 36% of their micro-blogs contents included celebrities’ images of wearing Burberry product. Chanel used 60% of their micro-blogs to broadcast and introduce their brand events. There was no sale information posted on Chanel and Burberry’s micro-blogs account. Through sentiment analysis, researchers also found the brands have very positive electronic word of mouth (e-WOM). Particular, luxury fashion brands are having better e-WOM than fast fashion brands.

본 논문은 기계 학습 기법 중에서 메모리 기반 학습을 사용하여 범용의 학습 가능한 한국어 문장 경계 인식기를 제안한다. 제안한 방법은 메모리 기반 학습 알고리즘 중 최 근린 이웃(kNN) 알고리즘을 사용하였으며, 이웃들을 이용한 문장 경계 결정을 위한 스 코어 값 계산을 위한 다양한 가중치 방법을 적용하여 이들을 비교 분석하였다. 문장 경 계 구분을 위한 자질로는 특정 언어나 장르에 제한적이지 않고 범용으로 적용될 수 있 는 자질만을 사용하였다. 성능 실험을 위하여 ETRI 코퍼스와 KAIST 코퍼스를 사용하 였으며, 성능 척도로는 정확도와 재현율이 사용되었다. 실험 결과 제안한 방법은 적은 학습 코퍼스만으로도 98.82%의 문장 정확률과 99.09%의 문장 재현율을 보였다.

In this study, the prediction technology of Hydrological Quantitative Precipitation Forecast (HQPF) was improved by optimizing the weather predictors used as input data for machine learning. Results comparison was conducted using bias and Root Mean Square Error (RMSE), which are predictive accuracy verification indicators, based on the heavy rain case on August 21, 2021. By comparing the rainfall simulated using the improved HQPF and the observed accumulated rainfall, it was revealed that all HQPFs (conventional HQPF and improved HQPF 1 and HQPF 2) showed a decrease in rainfall as the lead time increased for the entire grid region. Hence, the difference from the observed rainfall increased. In the accumulated rainfall evaluation due to the reduction of input factors, compared to the existing HQPF, improved HQPF 1 and 2 predicted a larger accumulated rainfall. Furthermore, HQPF 2 used the lowest number of input factors and simulated more accumulated rainfall than that projected by conventional HQPF and HQPF 1. By improving the performance of conventional machine learning despite using lesser variables, the preprocessing period and model execution time can be reduced, thereby contributing to model optimization. As an additional advanced method of HQPF 1 and 2 mentioned above, a simulated analysis of the Local ENsemble prediction System (LENS) ensemble member and low pressure, one of the observed meteorological factors, was analyzed. Based on the results of this study, if we select for the positively performing ensemble members based on the heavy rain characteristics of Korea or apply additional weights differently for each ensemble member, the prediction accuracy is expected to increase.

Heatwaves can affect human health and vegetation growth and bring about energy problems and socioeconomic damages, so the analysis and prediction of the heatwave is a crucial issue under a warming climate. This paper examines the production of STCI (Standard Temperature Condition Index) using ASOS (Automated Synoptic Observing System) in-situ observation data for the period of 1979-2020, and an STCI predictability assessment with an RF (Random Forest) model using ERA5 (ECMWF Reanalysis 5) meteorological variables. The accuracy was quite high with the MAE (Mean Absolute Error) of 0.365 and the CC (Correlation Coefficient) of 0.873, which corresponded to 7% to 10% difference for the range of STCI<1.5, and to 1% to 3% difference for the range of STCI>1.5, in terms of the probability density function. Also, we produced gridded maps for the summer STCI from 1979 to 2020 by utilizing the ERA5 raster data for the RF prediction model, which enables the spatial expansion of the ASOS point-based STCI to a continuous grid nationwide. The proposed method can be applied to forecasting of STCI by adopting future meteorological or climatic datasets.

학교소속감이 중도탈락을 막을 수 있는 중요한 변수임에도 불구하고 특히 고등학교 학생에 대한 학교소속감 연구는 그 수가 많지 않다. 아직 충분한 연구가 이루어지지 않은 분야에서 기계학습 기법을 활용하는 탐색적 연구가 학문적 기여를 할 수 있다. 본 연구의 주요 목적은 일반고 및 특성화고 학생들의 학교소속감을 예측하는 중요한 변수를 파악하는 것이다. 이를 위하여 부산교육종단 연구 2016 4차년도 데이터의 일반고와 특성화고 학생, 교사, 교장, 학교 변수를 모두 활용하였다. 구체적으로 75개 일반고 1,775명의 824개 변수 자료와 36개 특성화고 739명의 854개 변수 자료를 기계학습 기법으로 분석한 결과, 일반고와 특성화고에서 각각 20개와 21개의 학생, 교사, 학교 관련 변수가 선택되었다. 학교소속감을 개인의 심리적 차원에 초점을 맞추어 분석한 선행연구와 달리, 본 연구는 교사, 교장, 학교 변수까지 모두 모형에 투입함으로써 학교 현장에서의 변화를 꾀하였다. 기계학습 기법 중 벌점회귀모형으로 분류되는 glmmLasso를 활용하여 변수 선택 시 자료의 위계적 구조를 반영한 점 또한 연구 의의라 하겠다. 특히 특성화고 자료는 사례 수보다 변수 수가 더 많은 고차원 자료였으므로 기계학습 기법을 활용하는 것이 필수적이었다. 연구 결과를 토대로 고등학생의 학교소속감을 향상시키기 위하여 필요한 정책적 제언을 제시하고, 후속 연구주제 또한 논하였다.

Because of the population growth and industrialization in recent decades, the air quality over the world has been worsened with the increase of PM10 concentration. Korea is located near the eastern part of China which has many industrial complexes, so the consideration of China’s air quality is necessary for the PM10 prediction in Korea. This paper examines a machine learning-based modeling of the prediction of tomorrow’s PM10 concentration in the form of a gridded map using the AirKorea observations, Chinese cities’ air quality index, and NWP (numerical weather prediction) model data. A blind test using 23,048 cases in 2019 produced a correlation coefficient of 0.973 and an MAE (mean absolute error) of 4.097㎍/㎥, which is high accuracy due to the appropriate selection of input variables and the optimization of the machine learning model. Also, the prediction model showed stable predictability irrespective of the season and the level of PM10. It is expected that the proposed model can be applied to an operative system if a fine-tuning process using a larger database is accomplished.

In meteorological data, various studies are being conducted to improve the prediction performance of rainfall with irregular patterns, unlike temperature and solar radiation with certain patterns. Especially in the case of the short-term forecast model for Dong-Nae Forecasts provided by the Korea Meteorological Administration (KMA), forecast data are provided at 6-hour intervals, and there is a limit to analyzing the impact of disasters. In this study, Hydrological Quantitative Precipitation Forecast (HQPF) information was generated by applying the machine learning method to Local ENsemble prediction system (LENS), Radar-AWS Rainrates (RAR), AWS and ASOS observation data and Dong-Nae Forecast provided by the KMA. Through the preprocessing process, the temporal and spatial resolutions of all the data were converted to the same resolution, and the predictor of machine learning was derived through the factor analysis of the predictor. Considering the processing speed and expandability, the XGBoost method of machine learning was applied, and the Probability Matching (PM) method was applied to improve the prediction accuracy of heavy rainfall. As a result of evaluating the HQPF performance produced for 14 heavy rainfall events that occurred in 2020, it was found that the predicted performance of HQPF was improved quantitatively and qualitatively.

This paper studied the collision detection of robot manipulators for safe collaboration in human-robot interaction. Based on sensor-based collision detection, external torque is detached from subtracting robot dynamics. To detect collision using joint torque sensor data, a comparative study was conducted using data-based machine learning algorithm. Data was collected from the actual 3 degree-of-freedom (DOF) robot manipulator, and the data was labeled by threshold and handwork. Using support vector machine (SVM), decision tree and k-nearest neighbors KNN method, we derive the optimal parameters of each algorithm and compare the collision classification performance. The simulation results are analyzed for each method, and we confirmed that by an optimal collision status detection model with high prediction accuracy.

Water resources planning and management are, more and more, becoming important issue for water use and flood control due to the population increase, urbanization, and climate change. In particular, the estimating and the forecasting inflow of dam is the most important hydrologic issue for flood control and reliable water supply. Therefore, this study forecasted monthly inflow of Soyang river dam using VARMA model and 3 machine learning models. The forecasting models were constructed using monthly inflow data in the period of 1974 to 2016 and then the inflows were forecasted at 12- and 24-month ahead lead times. As a result, the forecasted monthly inflows by the models mostly were less than the observed ones, but the peak time and the variation pattern were well forecasted. Especially, the VARMA model showed very good performance in the forecasting. Therefore, the result of this study indicates that the VARMA model can be used efficiently to forecast hydrologic data and also used to establish water supply and management plan.

In this study, we compared the prediction performances according to the bias and dispersion of temperature using ensemble machine learning. Ensemble machine learning is meta-algorithm that combines several base learners into one prediction model in order to improve prediction. Multiple linear regression, ridge regression, LASSO (Least Absolute Shrinkage and Selection Operator; Tibshirani, 1996) and nonnegative ride and LASSO were used as base learners. Super learner (van der Lann et al ., 1997) was used to produce one optimal predictive model. The simulation and real data for temperature were used to compare the prediction skill of machine learning. The results showed that the prediction performances were different according to the characteristics of bias and dispersion and the prediction error was more improved in temperature with bias compared to dispersion. Also, ensemble machine learning method showed similar prediction performances in comparison to the base learners and showed better prediction skills than the ensemble mean.

The widespread sensors in a structural monitoring system provide vital support to its operation. Data is obtainedf rom sensors in a structural health monitoring system for integrity assessment of the structure, and false alarm will be frequently triggered if a faulty sensor is detected. In this study, a proposed method based on machine learning algorithm and Gaussian distribution is present to identify sensor fault.

For the purposes of enhancing usability of Numerical Weather Prediction (NWP), the quantitative precipitation prediction scheme by machine learning has been proposed. In this study, heavy rainfall was corrected for by utilizing rainfall predictors from LENS and Radar from 2017 to 2018, as well as machine learning tools LightGBM and XGBoost. The results were analyzed using Mean Absolute Error (MAE), Normalized Peak Error (NPE), and Peak Timing Error (PTE) for rainfall corrected through machine learning. Machine learning results (i.e. using LightGBM and XGBoost) showed improvements in the overall correction of rainfall and maximum rainfall compared to LENS. For example, the MAE of case 5 was found to be 24.252 using LENS, 11.564 using LightGBM, and 11.693 using XGBoost, showing excellent error improvement in machine learning results. This rainfall correction technique can provide hydrologically meaningful rainfall information such as predictions of flooding. Future research on the interpretation of various hydrologic processes using machine learning is necessary.

한국형 e-Navigation의 내항성 안전 모듈은 운항 중인 선박을 실시간으로 모니터링하고 내항성의 이상 상태를 사전에 경고함으로써 선박의 안정성을 확보하는 선내 원격 모니터링 서비스 중 하나이다. 일반적으로 선박설계를 위한 내항성능은 주어진 조건에서 선체 운동 시뮬레이션을 수행하여 평가하여 왔다. 하지만 운항 중 선박의 내항성능을 실시간으로 평가하기 위해 이러한 시뮬레이션을 실제 운항조건에 맞추어 수행하는 것은 계산시간의 한계로 인해 현실적이지 않다. 본 연구에서는 기계학습 기반의 근사모델을 활용하여 선박의 내항성능 평가 요소들 중 하나인 횡동요 운동특성을 합리적으로 보다 빠르게 예측하는 방법을 소개하고자 한다. 다양한 학습 기법과 데이터의 샘플링 조건을 적용하여, 얻어진 근사모델의 결과와 운동해석 결과의 오차가 거의 1% 내로 일치함을 보였다. 따라서 이러한 방법을 활용하면 선박의 실시간 내항 성능을 평가하는데 효율적으로 사용할 수 있을 것으로 판단된다.

Arctic sea ice as an indicator of climate change plays an important role in controlling global climate system. Thus, accurate observation and prediction of Sea Ice Concentration (SIC) is essential for understanding global climate change. In this study, we aim to improve the prediction accuracy of SIC by using machine learning and Regional Climate Model (RCM) data for a more robust method and a higher spatial resolution. Using the CORDEX RCM and NASA SIC data between January 1981 and December 2015, we developed three statistical models using Multiple Linear Regression (MLR), Support Vector Machine (SVM), and Deep Neural Network (DNN) which can deal with the non-linearity problem, respectively. The DNN model showed the best performance among the three models with the significant correlation between the predictive and observed SIC (r=0.811, p-value < 0.01)and the Root Mean Square Error (RMSE) of 0.258. With deeper considerations of the polar fronts and the characteristics of ocean current and tide, the DNN model can be applied for near future prediction of Arctic sea ice changes.

Solar radiation forecasts are important for predicting the amount of ice on road and the potential solar energy. In an attempt to improve solar radiation predictability in Jeju, we conducted machine learning with various data mining techniques such as tree models, conditional inference tree, random forest, support vector machines and logistic regression. To validate machine learning models, the results from the simulation was compared with the solar radiation data observed over Jeju observation site. According to the model assesment, it can be seen that the solar radiation prediction using random forest is the most effective method. The error rate proposed by random forest data mining is 17%.

본 연구는 기계학습적 접근법인 LASSO 기법을 우리나라 TIMSS 2011 중학교 2학년 자료 에 적용하였다. TIMSS의 100개의 설명변수를 모형에 모두 투입하여 22개 변수를 선택하였을 때, 이 모형의 예측정확도는 약 80%였다. 학생의 수학적 자기효능감, 수학에 대한 태도, 어머 니의 교육 수준, 그리고 가정 보유 장서 수와 같은 가정의 교육자원 변수가 학생의 수학 성취 수준에 영향을 미치는 것으로 나타났으며, 이는 기존 연구 결과와 일치하였다. 본 연구에서 학 생의 수학 성취수준과 관련과 있다고 새롭게 탐색된 변수로 수학숙제 시간, 학생의 과학적 자 기효능감, 과학숙제 부여 빈도 등이 있었다. 연구 함의 및 향후 연구 주제 또한 논의되었다.

A novel disaggregation model that combines a machine learning model and kriging of residuals is presented to map precipitation at a fine scale from coarse scale precipitation data. Random forest (RF) and fine scale auxiliary variables are used to estimate trend components at a fine scale. Residual components are then estimated by area-to-point residual kriging. A case study of spatial disaggregation of TRMM monthly precipitation data acquired over the Korean peninsula is carried out to illustrate the potential of the presented disaggregation method. From the evaluation results, the presented method outperformed the RF-based disaggregation method that only considers trend components and ignores residual components, in terms of accuracy statistics and the ability of coherent predictions. This case study indicates that accounting for residual components by applying a proper spatial prediction method such as area-to-point kriging is very important in spatial disaggregation of coarse scale spatial data, even though advanced regression models such as RF could have high goodness of fit for the quantification of relationships between a target attribute and auxiliary variables.

본 논문은 기계학습 기법을 이용한 게임 캐릭터를 제어하는 엔진을 설계하고 구현하는 것을 제안한다. 제안된 엔진은 실제 게임에서 상황 데이터를 추출하여 지식 데이터로 사용하므로 지능 캐릭터의 행동 패턴을 게이머들이 쉽게 인식하지 못하는 장점이 있다. 이를 위하여 상황 데이터를 추출하여 학습하는 모듈과 임의의 상황 데이터에 대하여 최적의 상황 제어를 판단하는 시험 모듈을 개발하는 것을 제안하였다. 구현된 엔진은 FEAR에 이식되고 Quake2 게임에 적용되었다. 또한 개발된 엔진의 올바른 동작과 효율성을 위하여 다양한 실험을 하였다. 실험으로부터 개발된 엔진이 올바르게 동작할 뿐만 아니라 제한된 시간 내에 효율적으로 동작함을 알 수 있었다.