실제세계에서 데이터 수집의 비용과 한계를 고려할 때, 시뮬레이션 생성 환경은 데이터 생성 과 다양한 시도에 있어 효율적인 대안이다. 이 연구에서는 Unity ML Agent를 로그라이크 장 르에 적합한 강화학습 모델로 구현하였다. 간단한 게임에Agent를 이식하고, 이 Agent가 적을 인식하고 대응하는 과정을 코드로 작성하였다. 초기 모델은 조준사격의 한계를 보였으나 RayPerceptionSensor-Component2D를 통해 Agent의 센서 정보를 직접 제공함으로써, Agent가 적을 감지하고 조준 사격을 하는 능력을 관찰할 수 있었다. 결과적으로, 개선된 모델 은 평균3.81배 향상된 성능을 보여주었으며, 이는 Unity ML Agent가 로그라이크 장르에서 강화학습을 통한 데이터 수집이 가능함을 입증한다.

The government is implementing a policy to expand eco-friendly energy as a power source. However, the output of new and renewable energy is not constant. It is difficult to stably adjust the power supply to the power demand in the power system. Therefore, the government predicts day-ahead the amount of renewable energy generation to cope with the output volatility caused by the expansion of renewable energy. It is a system that pays a settlement amount if it transitions within a certain error rate the next day. In this paper, Machine Learning was used to study the prediction of power generation within the error rate.

PURPOSES : Construction cost estimates are important information for business feasibility analysis in the planning stage of road construction projects. The quality of current construction cost estimates are highly dependent on the expert's personal experience and skills to estimate the arithmetic average construction cost based on past cases, which makes construction cost estimates subjective and unreliable. An objective approach in construction cost estimation shall be developed with the use of machine learning. In this study, past cases of road projects were analyzed and a machine learning model was developed to produce a more accurate and time-efficient construction cost estimate in teh planning stage. METHODS : After conducting case analysis of 100 road construction, a database was constructed including the road construction's details, drawings, and completion reports. To improve the construction cost estimation, Mallow's Cp. BIC, Adjusted R methodology was applied to find the optimal variables. Consequently, a plannigs-stage road construction cost estimation model was developed by applying multiple regression analysis, regression tree, case-based inference model, and artificial neural network (ANN, DNN). RESULTS : The construction cost estimation model showed excellent prediction performance despite an insufficient amount of learning data. Ten cases were randomly selected from the data base and each developed machine learning model was applied to the selected cases to calculate for the error rate, which should be less than 30% to be considered as acceptable according to American Estimating Association. As a result of the analysis, the error rates of all developed machine learning models were found to be acceptable with values rangine from 17.3% to 26.0%. Among the developed models, the ANN model yielded the least error rate. CONCLUSIONS : The results of this study can help raise awareness of the importance of building a systematic database in the construction industry, which is disadvantageous in machine learning and artificial intelligence development. In addition, it is believed that it can provide basic data for research to determine the feasibility of construction projects that require a large budget, such as road projects.

인공지능은 4차 산업혁명의 프레임이 소개된 이후 점차 보편적인 기술로 자리를 잡아가고 있으며, 인공지능 관련 특허 출원도 크게 증가하고 있다. 최근에는 특허 생태계가 출원 건수 위주의 양적 경쟁에서 고품질의 특허 확보라는 질적 경쟁으로 패러다임이 변화되면서, 저품질 특허로 인한 비용 손실에 관심이 높아지고 있다. 이러한 배경으로 본 연구에서는 머신러닝과 Doc2Vec 알고리즘을 활용하여 특허 품질을 예측하는 방법을 제안하고자 한다. 본 연구를 위해 WIPO에서 정의한 CPC 코드를 활용하여 미국 특허청(USPTO)에 등록된 인공지능 관련 특허 데이터를 추출하였고, 이를 통해 정형 데이터 기반 19개 변수, 비정형 데이터 기반 7개 변수를 개발하였다. 특히, 새롭게 제안하는 Doc2Vec 알고리즘을 이용한 제목과 초록 텍스트 유사도 변수는 고품질 특허를 예측하는데 영향을 미칠 것으로 판단된다. 이에 유사도 변수의 효과를 확인하기 위해 유사도 변수를 포함한 앙상블 기반 머신러닝 모델과 포함하지 않은 모델을 개발하여 비교하였다. 실험 결과, 유사도 변수를 포함한 모델이 AUC 0.013, f1-score 0.025가 높게 나타나 더 우수한 성능을 보였다. 이는 유사도 변수가 고품질 특허 예측에 기여한다는 것을 시사한다. 또한, SHAP을 이용하여 블랙박스 형태의 머신러닝 변수 영향도를 설명하였다. 본 연구를 통해 핵심 기술 분야인 인공지능과 같은 영역에서 특허의 품질을 예측하고, 고품질 특허 개발을 장려함으로써 사회적 가치를 실현하는 데 기여할 수 있을 것으로 기대한다.

Fouling is an inevitable problem in membrane water treatment plant. It can be measured by trans-membrane pressure (TMP) in the constant flux operation, and chemical cleaning is carried out when TMP reaches a critical value. An early fouilng alarm is defined as warning the critical TMP value appearance in advance. The alarming method was developed using one of machine learning algorithms, decision tree, and applied to a ceramic microfiltration (MF) pilot plant. First, the decision tree model that classifies the normal/abnormal state of the filtration cycle of the ceramic MF pilot plant was developed and it was then used to make the early fouling alarm method. The accuracy of the classification model was up to 96.2% and the time for the early warning was when abnormal cycles occurred three times in a row. The early fouling alram can expect reaching a limit TMP in advance (e.g., 15-174 hours). By adopting TMP increasing rate and backwash efficiency as machine learning variables, the model accuracy and the reliability of the early fouling alarm method were increased, respectively.

Nowadays, artificial intelligence model approaches such as machine and deep learning have been widely used to predict variations of water quality in various freshwater bodies. In particular, many researchers have tried to predict the occurrence of cyanobacterial blooms in inland water, which pose a threat to human health and aquatic ecosystems. Therefore, the objective of this study were to: 1) review studies on the application of machine learning models for predicting the occurrence of cyanobacterial blooms and its metabolites and 2) prospect for future study on the prediction of cyanobacteria by machine learning models including deep learning. In this study, a systematic literature search and review were conducted using SCOPUS, which is Elsevier’s abstract and citation database. The key results showed that deep learning models were usually used to predict cyanobacterial cells, while machine learning models focused on predicting cyanobacterial metabolites such as concentrations of microcystin, geosmin, and 2-methylisoborneol (2-MIB) in reservoirs. There was a distinct difference in the use of input variables to predict cyanobacterial cells and metabolites. The application of deep learning models through the construction of big data may be encouraged to build accurate models to predict cyanobacterial metabolites.

목적 : 4차 산업혁명이 진행됨에 따라 타각적 굴절검사값, 수차 및 동공크기 등을 이용하여 최적의 안경처방값 을 도출해주는 머신러닝(machine learning)을 개발하고자 하였다. 방법: 시력에 영향을 줄 수 있는 안질환 및 전신질환이 없고 안구 수술 이력이 없는 근시안(1,000안)을 대상으로 진행하였다. I-Profilerplus(Zeiss, Berlin, Germany)를 사용하여 타각적 굴절이상도(objective-refraction) 및 안구수차(ocular wavefront-aberration), 동공 크기를 측정하였고, 자각적 굴절이상도(subjective-refraction)는 Visuphor500(Zeiss, Berlin, Germany)를 사용하여 구면 굴절력(S, Diopter), 원주 굴절력(C, Diopter), 난시 축(Ax, °)을 측정하였다. 측정 후, 파이썬(Python, version 3.10)을 이용하여 머신러닝 모델 생성 및 예측 성능을 확인하였다. 결과: 자각적 굴절이상도에서 구면 굴절력에 영향을 미치는 요인은 타각적 구면 굴절력, defocus aberration, spherical aberration, trefoil aberration 순으로 높았고, 원주 굴절력에 영향을 미치는 요인은 타각적 원주 굴 절력, defocus aberration, coma aberration, trefoil aberration 순으로 높았으며, 난시 축은 타각적 난시축만 영향을 미치는 것으로 나타났다. 구면 굴절력, 원주 굴절력, 난시 축의 자각적 굴절이상도와 머신러닝 예상값은 차이가 없는 것으로 나타났다(p=0.976, 0.948, and 0.349, respectively). 결론 : 자각적 굴절이상도를 예측하는 머신러닝 모델을 생성하였고, 해당 모델의 예측된 값과 자각적 굴절이상 도와 유의한 차이가 없는 것을 통해 예측 정확도를 확인하였으며 앞으로 개인 맞춤형 처방을 위한 정확한 안경처 방값을 도출하는데 기초자료가 될 수 있을 것으로 생각된다.

The entire industry is increasing the use of big data analysis using artificial intelligence technology due to the Fourth Industrial Revolution. The value of big data is increasing, and the same is true of the production technology. However, small and medium -sized manufacturers with small size are difficult to use for work due to lack of data management ability, and it is difficult to enter smart factories. Therefore, to help small and medium -sized manufacturing companies use big data, we will predict the gross production time through machine learning. In previous studies, machine learning was conducted as a time and quantity factor for production, and the excellence of the ExtraTree Algorithm was confirmed by predicting gross product time. In this study, the worker's proficiency factors were added to the time and quantity factors necessary for production, and the prediction rate of LightGBM Algorithm knowing was the highest. The results of the study will help to enhance the company's competitiveness and enhance the competitiveness of the company by identifying the possibility of data utilization of the MES system and supporting systematic production schedule management.

Gate valves are hydraulic components used to shut-off the water flow in water distribution systems. Gate valves may fail owing to various aspects such as leakage through seats, wearing of packing, and corrosion. Because it is considerably challenging to detect valve malfunctioning until the operator identifies a significant fault, failure of the gate valve may lead to a severe accident event associated with water distribution systems. In this study, we proposed a methodology to diagnose the faults of gate valves. To measure the pressure difference across a gate valve, two pressure transducers were installed before and after the gate valve in a pilot-scaled water distribution system. The obtained time-series pressure difference data were analyzed using a machine learning algorithm to diagnose faults. The validation of whether the flow rate of the pipeline can be predicted based on the pressure difference between the upstream and downstream sides of the valve was also performed.

With about 80% of the global economy expected to shift to the global market by 2030, exports of reverse direct purchase products, in which foreign consumers purchase products from online shopping malls in Korea, are growing 55% annually. As of 2021, sales of reverse direct purchases in South Korea increased 50.6% from the previous year, surpassing 40 million. In order for domestic SMEs(Small and medium sized enterprises) to enter overseas markets, it is important to come up with export strategies based on various market analysis information, but for domestic small and medium-sized sellers, entry barriers are high, such as lack of information on overseas markets and difficulty in selecting local preferred products and determining competitive sales prices. This study develops an AI-based product recommendation and sales price estimation model to collect and analyze global shopping malls and product trends to provide marketing information that presents promising and appropriate product sales prices to small and medium-sized sellers who have difficulty collecting global market information. The product recommendation model is based on the LTR (Learning To Rank) methodology. As a result of comparing performance with nDCG, the Pair-wise-based XGBoost-LambdaMART Model was measured to be excellent. The sales price estimation model uses a regression algorithm. According to the R-Squared value, the Light Gradient Boosting Machine performs best in this model.

본 연구에서는 2 016년부터 2 02 0년까지 내륙 관측소 중 안개 최다발 지역인 안동을 대상으로 XGBoost-DART 머신러닝 알고리즘을 이용하여 1 시간 후 안개 유무를 예측하였다. 기상자료, 농업관측자료, 추가 파생자료와 각 자료 를 오버 샘플링한 확장자료, 총 6개의 데이터 세트를 사용하였다. 목측으로 획득한 기상현상번호와 시정계 관측으로 측 정된 시정거리 자료를 각각 안개 유[1]무[0]로 이진 범주화하였다. 총 12개의 머신러닝 모델링 실험을 설계하였고, 안개 가 사회와 지역사회에 미치는 유해성을 고려하여 모델의 성능은 재현율과 AUC-ROC를 중심으로 평가하였다. 전체적으 로, 오버샘플링한 기상자료와 기상현상번호 기반의 예측 목표를 조합한 실험이 최고 성능을 보였다. 이 연구 결과는 머 신러닝 알고리즘을 활용한 안개 예측에 있어서, 목측으로 획득한 기상현상번호의 중요성을 암시한다.

Research and interest in sustainable printing are increasing in the packaging printing industry. Currently, predicting the amount of ink required for each work is based on the experience and intuition of field workers. Suppose the amount of ink produced is more than necessary. In this case, the rest of the ink cannot be reused and is discarded, adversely affecting the company's productivity and environment. Nowadays, machine learning models can be used to figure out this problem. This study compares the ink usage prediction machine learning models. A simple linear regression model, Multiple Regression Analysis, cannot reflect the nonlinear relationship between the variables required for packaging printing, so there is a limit to accurately predicting the amount of ink needed. This study has established various prediction models which are based on CART (Classification and Regression Tree), such as Decision Tree, Random Forest, Gradient Boosting Machine, and XGBoost. The accuracy of the models is determined by the K-fold cross-validation. Error metrics such as root mean squared error, mean absolute error, and R-squared are employed to evaluate estimation models' correctness. Among these models, XGBoost model has the highest prediction accuracy and can reduce 2134 (g) of wasted ink for each work. Thus, this study motivates machine learning's potential to help advance productivity and protect the environment.

본 연구의 주요 목적은 회귀기반의 다양한 머신러닝 알고리즘을 개발하고 다양한 농업 분야에서 사용되는 트랙터의 연료 소비량을 예측하는 것이다. 비포장 도로주행 농업 기계중에서도 사용 비중이 가장 높은 트랙터를 선정하였다. 실제 농가에 방문하여 현업 전문가 조언을 바탕으로 연구하여 설문지를 작성하였으며, 설문 대상은 경남 사천시에 있는 농가 10곳, 진주시에 있는 농가 62곳 등, 총 72곳의 농가이다. 농작업으로는 벼농사, 보리농사, 밭농사 등이 있으며, 작업내용으로는 쟁기, 로터리, 비료살포, 베토, 모내기작업 등이 있다. 다중 회귀분석을 통해 연료 소비량 예측에 영향을 미치는 변수(마력, 기계사용연수, 경작면적, 작업 시간)를 추출하였고. 머신러닝 회귀 학습기 모형으로 학습하여 예측 모형의 성능을 검증하였다. 연료 소비량을 예측하는 모델의 성능은 결정 계수(R), RMSE (제곱 평균 제곱근 오차), MSE (평균 제곱 오차) 및 MAE (평균 절대 오차)를 포함한 4가지 통계적 품질 매개변수를 사용하여 결정되었다. 연구 결과 4가지 모델(다중회귀, 랜덤포레스트, 아다부스트, K-최근접 이웃) 중 K-최근접 이웃의 성능이 제일 높은 것으로 나타났다. 결론적으로 본 연구의 결과는 실제 농가의 연료 소비량을 예측하여 면세유 유통의 투명성을 확보하고 추후 개발 모델의 의사결정에 활용될 수 있을 것으로 기대된다.

해양사고 발생시 실종자는 해양에 노출된 시간이 길어질수록 생존확률이 빠르게 감소하기 때문에 인명구조를 위해서는 신 속한 수색이 필요하다. 또한 해양의 수색영역은 육상에 비해서 매우 넓기 때문에 효율적인 수색을 위해서는 선박을 이용한 육안수색보 다는 인공위성이나 항공기에 탑재된 센서를 이용한 해상 객체 탐지 기술의 적용이 필요하다. 본 연구는 항공기에 탑재된 초분광 영상 센서를 이용하여 해양에서 객체를 신속하게 탐지하기 위한 목적으로 진행되었다. 초분광 영상 센서로 촬영된 영상은 8,241 × 1,024의 공간 해상도를 가지며, 한 화소당 0.7 m의 분해능과 127개의 스펙트럼으로 구성된 대용량의 데이터이다. 본 연구에서는 이러한 데이터 를 신속하게 분석하기 위한 목적으로 DBSCAN을 사용한 해수 식별 알고리즘과 밀도 기반의 육지 제거 알고리즘을 결합한 해상 객체 탐지 모델을 개발하였다. 개발한 모델은 초분광 영상에 적용하였을 때 약 5 km2의 해상 영역을 100초 내로 분석할 수 있는 성능을 보 였다. 또한 개발한 모델의 탐지 정확도를 평가하기 위해서 항공기를 이용하여 목포, 군산, 여수 지역의 초분광 영상을 촬영하였으며, 본 연구에서 개발한 탐지 모델에 적용한 결과, 실험 영상 내의 선박들을 90 %의 정확도로 탐지할 수 있는 결과를 얻었다. 본 연구에서 개발된 기술은 소형 선박의 수색·구조 활동을 지원하는 중요한 정보로 활용될 수 있을 것으로 기대한다.

The management of algal bloom is essential for the proper management of water supply systems and to maintain the safety of drinking water. Chlorophyll-a(Chl-a) is a commonly used indicator to represent the algal concentration. In recent years, advanced machine learning models have been increasingly used to predict Chl-a in freshwater systems. Machine learning models show good performance in various fields, while the process of model development requires considerable labor and time by experts. Automated machine learning(auto ML) is an emerging field of machine learning study. Auto ML is used to develop machine learning models while minimizing the time and labor required in the model development process. This study developed an auto ML to predict Chl-a using auto sklearn, one of most widely used open source auto ML algorithms. The model performance was compared with other two popular ensemble machine learning models, random forest(RF) and XGBoost(XGB). The model performance was evaluated using three indices, root mean squared error, root mean squared error-observation standard deviation ratio(RSR) and Nash-Sutcliffe coefficient of efficiency. The RSR of auto ML, RF, and XGB were 0.659, 0.684 and 0.638, respectively. The results shows that auto ML outperforms RF, and XGB shows better prediction performance than auto ML, while the differences between model performances were not significant. Shapley value analysis, an explainable machine learning algorithm, was used to provide quantitative interpretation about the model prediction of auto ML developed in this study. The results of this study present the possible applicability of auto ML for the prediction of water quality.

A reliable prediction model of national occupational accident fatality rate can be used to evaluate level of safety and health protection for workers in a country. Moreover, the socio-economic aspects of occupational accidents can be identified through interpretation of a well-organized prediction model. In this paper, we propose a machine learning based relative comparison methods to predict and interpret a national occupational accident fatality rate based on socio-economic indicators. First, we collected 29 years of the relevant data from 11 developed countries. Second, we applied 4 types of machine learning regression models and evaluate their performance. Third, we interpret the contribution of each input variable using Shapley Additive Explanations(SHAP). As a result, Gradient Boosting Regressor showed the best predictive performance. We found that different patterns exist across countries in accordance with different socio-economic variables and occupational accident fatality rate.