• Journal of the Institute of Electronics and Information Engineers
• /
• v.53 no.1
• /
• pp.71-78
• /
• 2016

This paper proposes the new electric power demand forecast model which is based on an artificial neural network and considers time and weather factors. Time factors are selected by measuring the autocorrelation coefficients of load demand in summer and winter seasons. Weather factors are selected by using Pearson correlation coefficient The important weather factors are temperature and dew point because the correlation coefficients between these factors and load demand are much higher than those of the other factors such as humidities, air pressures and wind speeds. The experimental results show that the proposed model using time and seasonal weather factors improves the load demand forecasts to a great extent.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2017.07a
• /
• pp.394-396
• /
• 2017

본 논문에서는 최근이슈가 되고 있는 인공지능에 대한 많은 기술 가운데 인공신경망을 활용하여 자신이 가고자 하는곳의 여행정보를 스케줄링 하는 시스템을 제안한다. 인공신경망 중에서도 비지도 학습(unsupervised learning)방식을 이용하며 이용자의 가중치에 따라 여행의 나이, 기간, 장소, 종류, 날씨, 계절, 인원 등으로 여행에서의 요소들을 히든레이어로 구성하여 여행지의 스케줄을 구성하여 이용자에게 제공하는 형태이다. 가중치에 따른 여행지의 분류작업이 완료가 되면 기간과 장소의 위치정보에 따라 스케줄링 작업을 완료하게 된다. 기존의 여행지에 대한 정보를 검색에 의해서 이루어지던 환경에서 인공신경망을 활용하여 원하는 여행정보를 추출함으로써 이용자에게 여행정보에 대한 체계화된 정보를 제공할 수 있다.

• Journal of the Korean Data Analysis Society
• /
• v.20 no.6
• /
• pp.2933-2940
• /
• 2018

It is known that fruit is more affected by the weather than other crops. Therefore, in order to create high value for farmers, it is necessary to develop a wholesale price model considering the weather. Peaches produced under relatively limited conditions were chosen as subjects of study. The data were collected from 2015 to 2017 provided by okdab 4.0. The meteorological data used for the analysis were generated by weighting the cultivation area and the variables with high correlation among the weather data were selected from the day before to 7 days before. Randomforest, gradient boosting machine, and XGboost were used for the analysis. As a result of analysis, XGboost showed the best performance in the sense of RMSE and correlation, and price prediction was comparatively well predicted, but the accuracy of the trading volume prediction was not so good enough. The top three weather variables affecting to the peach were minimum temperature, average maximum temperature, and precipitation.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2017.04a
• /
• pp.708-711
• /
• 2017

최근 의류업계에서는 데이터마이닝을 이용하여 의상을 추천하는 시스템에 대한 연구가 활발하게 진행되고 있다. 하지만 기존 연구들은, 의상구매가 온 오프라인 모두에서 활발함에도 불구하고 온라인 쇼핑몰에서 얻을 수 있는 데이터에 국한되어 연구가 진행되고 있다. 본 논문에서는 온라인 데이터 위주의 기존 의상 추천 시스템을 스마트 홈 미러의 가상 착의시스템을 사용하여 온 오프라인 데이터를 모두 반영한 추천시스템을 구현했다. 또한 사용자에게 적합한 추천시스템을 제공하기 위해 지역별 인구분포와 사용자 기본DB를 단계별로 그룹화 했다. 정확도와 사용자 만족도를 향상 시키고자 단계별로 가중치를 부여해 협업 필터링과 날씨, 종류, 색상을 속성으로 한 내용기반 필터링을 결합하는 시스템을 제시했다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2019.05a
• /
• pp.330-333
• /
• 2019

본 연구는 기존 관중 수 예측에 주로 사용되는 ARIMA 모형과 다른 GKFN(Network with Gaussian kernel functions) 모델을 시계열 모델로 제안하고 여러 변수 간의 상관관계를 분석한 MLP(Multilayer Perceptron) 모델을 각각 따로 만들어 두 가지 RMSE값의 가중치를 결합한 새로운 모델을 최종적으로 제안한다. GKFN 모델은 phase space 분석을 위해 smoothness measure를 측정하고 커널 개수를 늘려가며 학습시키는 방법이다. 또한, MLP 모델은 관중 수에 영향을 주는 여러 변수(날짜, 날씨 등 팀과 관련된 특징들)의 상관관계를 correlation coefficient 값을 이용해 분석하고 높은 상관관계를 가지는 변수들을 이용해 MLP 모델을 만들어 학습하는 것이다. 이를 통해 프로야구팀 기아 타이거즈의 일일 단위 관중 수를 예측하고자 하였다. 관중 수 예측을 통해 구단과 관객 모두 긍정적인 활용이 가능할 것이다. 훈련 자료는 2010년부터 2018년까지 9년 동안 기아 타이거즈의 일별 관중 수를 자료로 하였다.

• Journal of the Korean Data and Information Science Society
• /
• v.28 no.5
• /
• pp.1125-1132
• /
• 2017

In a situation where the severity and frequency of drought events getting stronger and higher, many studies related to drought forecast have been conducted to improve the drought forecast accuracy. However it is difficult to predict drought events using a single model because of nonlinear and complicated characteristics of temporal behavior of drought events. In this study, in order to overcome the shortcomings of the single model approach, we first build various single models capable to explain the relationship between the meteorological drought index, Standardized Precipitation Index (SPI), and other independent variables such as world climate indices. Then, we developed a combined models using Stochastic Gradient Descent method among Ensemble Learnings.

• Journal of the Institute of Convergence Signal Processing
• /
• v.20 no.3
• /
• pp.138-144
• /
• 2019

In a modern society, people are concerned seriously about their travel destinations depending on time, economic problem. In this paper, we propose an non-hierarchical clustering based hybrid recommendation using context knowledge. The proposed method is personalized way of recommended knowledge about preferred travel places according to the user's location, place, and weather. Based on 14 attributes from the data collected through the survey, users with similar characteristics are grouped using a non-hierarchical clustering based hybrid recommendation. This makes more accurate recommendation by weighting implicit and explicit data. The users can be recommended a preferred travel destination without spending unnecessary time. The performance evaluation uses accuracy, recall, F-measure. The evaluation result was shown 0.636 accuracy, 0.723 recall, and 0.676 F-measure.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.33 no.4
• /
• pp.287-296
• /
• 2015

Recent navigation systems provide quick guide services, based on processing real-time traffic information and past traffic information by applying predictable pattern for traffic information. However, the current pattern for traffic information predicts traffic information by processing past information that it presents an inaccuracy problem in particular circumstances(accidents and weather). So, this study presented a more precise predictive traffic information system than historical traffic data first by analyzing route search data which the drivers ask in real time for the quickest way then by grasping traffic congestion levels of the route in which future drivers are supposed to locate. First results of this study, the congested route from Yang Jae to Mapo, the analysis result shows that the accuracy of the weighted value of speed of existing commonly congested road registered an error rate of 3km/h to 18km/h, however, after applying the real predictive traffic information of this study the error rate registered only 1km/h to 5km/h. Second, in terms of quality of route as compared to the existing route which allowed for an earlier arrival to the destination up to a maximum of 9 minutes and an average of up to 3 minutes that the reliability of predictable results has been secured. Third, new method allows for the prediction of congested levels and deduces results of route searches that avoid possibly congested routes and to reflect accurate real-time data in comparison with existing route searches. Therefore, this study enabled not only the predictable gathering of information regarding traffic density through route searches, but it also made real-time quick route searches based on this mechanism that convinced that this new method will contribute to diffusing future traffic flow.