• KIPS Transactions on Computer and Communication Systems
• /
• v.5 no.10
• /
• pp.293-302
• /
• 2016

Since the electricity is produced and consumed simultaneously, predicting the electric load and securing affordable electric power are necessary for reliable electric power supply. In particular, a university campus is one of the highest power consuming institutions and tends to have a wide variation of electric load depending on time and environment. For these reasons, an accurate electric load forecasting method that can predict power consumption in real-time is required for efficient power supply and management. Even though various influencing factors of power consumption have been discovered for the educational institutions by analyzing power consumption patterns and usage cases, further studies are required for the quantitative prediction of electric load. In this paper, we build an electric load forecasting model by implementing and evaluating various machine learning algorithms. To do that, we consider three building clusters in a campus and collect their power consumption every 15 minutes for more than one year. In the preprocessing, features are represented by considering periodic characteristic of the data and principal component analysis is performed for the features. In order to train the electric load forecasting model, we employ both artificial neural network and support vector machine. We evaluate the prediction performance of each forecasting model by 5-fold cross-validation and compare the prediction result to real electric load.

• Journal of Korea Water Resources Association
• /
• v.54 no.9
• /
• pp.667-680
• /
• 2021

We investigated changes in frost days and frost-free periods and to comparatively assess frost event prediction models developed using logistic regression (LR), random forest (RF), and long short-term memory (LSTM) networks. The meteorological variables for the model development were collected from the Suwon, Cheongju, and Gwangju stations for the period of 1973-2019 for spring (March - May) and fall (September - November). The developed models were then evaluated by Precision, Recall, and f-1 score and graphical evaluation methods such as AUC and reliability diagram. The results showed that significant decreases (significance level of 0.01) in the frequencies of frost days were at the three stations in both spring and fall. Overall, the evaluation metrics showed that the performance of RF was highest, while that of LSTM was lowest. Despite higher AUC values (above 0.9) were found at the three stations, reliability diagrams showed inconsistent reliability. A further study is suggested on the improvement of the predictability of both frost events and the first and last frost days by the frost event prediction models and reliability of the models. It would be beneficial to replicate this study at more stations in other regions.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.16 no.2
• /
• pp.98-103
• /
• 2002

In case of power demand forecasting, the most important problems are to deal with the load of special-days. Accordingly, this paper presents the method that forecasting long (the Lunar New Year, the Full Moon Festival) and short(the Planting Trees Day, the Memorial Day, etc) special-days peak load using neural networks and regression models. long and short special-days peak load forecast by neural networks models uses pattern conversion ratio and four-order orthogonal polynomials regression models. There are using that special-days peak load data during ten years(1985∼1994). In the result of special-days peak load forecasting, forecasting % error shows good results as about 1 ∼2[%] both neural networks models and four-order orthogonal polynomials regression models. Besides, from the result of analysis of adjusted coefficient of determination and F-test, the significance of the are convinced four-order orthogonal polynomials regression models. When the neural networks models are compared with the four-order orthogonal polynomials regression models at a view of the results of special-days peak load forecasting, the neural networks models which uses pattern conversion ratio are more effective on forecasting long special-days peak load. On the other hand, in case of forecasting short special-days peak load, both are valid.

• Journal of the Korea Institute of Building Construction
• /
• v.18 no.5
• /
• pp.455-461
• /
• 2018

The purpose of this study is to evaluate the tensile strength performance of the polyurethane coating material used as the waterproofing material in concrete structures. A linear regression equation is proposed to establish a correlation on the tensile strength of polyurethane coating membrane against periodic exposure to chemical degradation. The polyurethane film membrane showed a minimum strength of 23% to a maximum of 38% when subjected to chemical degradation. The elongation rate showed a relation with the tensile strength deterioration rate of at least 15% to 22% at maximum, and the proposed regression equation could be used to predict the degree of performance change of the polyurethane coating membrane under chemical degradation condition.

• Journal of the Korea Computer Graphics Society
• /
• v.23 no.4
• /
• pp.21-29
• /
• 2017

Previous visual analytics researches has focused on reducing the uncertainty of predicted results using a variety of interactive visual data exploration techniques. The main purpose of the interactive search technique is to reduce the quality difference of the predicted results according to the level of the decision maker by understanding the relationship between the variables and choosing the appropriate model to predict the unknown variables. However, it is difficult to create a predictive model which forecast time series data whose overall trends is unknown such as youth physical growth data. In this paper, we pro pose a novel predictive analysis technique to forecast the physical growth value in small pieces of time series data with un certain trends. This model estimates the distribution of data at a particular point in time. We also propose a visual analytics system that minimizes the possible uncertainties in predictive modeling process.

• Review of KIISC
• /
• v.26 no.2
• /
• pp.35-38
• /
• 2016

본 논문에서는 핀테크 보안에 활용 가능한 딥러닝 기술을 살펴본다. 먼저 인공지능과 관련된 보안 이슈를 인공지능이 사람을 위협하는 상황에 대한 보안(Security FROM AI), 인공지능 시스템이나 서비스를 악의적인 공격으로부터 보호하는 이슈(Security OF AI), 인공지능 기술을 활용해 보안 문제를 해결하는 것(Security BY AI) 3가지로 구분하여 살펴본다. Security BY AI의 일환으로 딥러닝에 기반한 비정상탐지(anomaly detection)과 회귀분석(regression)기법을 설명하고, 이상거래탐지, 바이오인증, 피싱, 파밍 탐지, 본인확인, 명의도용탐지, 거래 상대방 신뢰도 분석 등 핀테크 보안 문제에 활용할 수 있는 방안을 살펴본다.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2012.05a
• /
• pp.364-364
• /
• 2012

수자원 연구의 주요 목적인 효과적인 홍수 및 가뭄관리를 하기 위해서는 그 연구의 기초가 되는 자료를 관측하고 정도(accuracy, 精度)를 향상시키는 연구 또한 매우 중요한 부분이라고 볼 수 있다. 이러한 점에서 수위-유량측정의 경우, 관측자의 숙련도와 계측기 오차에 따라 관측값에 미치는 영향이 큰 특징을 갖고 있어 유량측정의 정확성을 높이고자 진보된 계측기의 개발 및 분석 방법에 관한 연구는 꾸준히 진행되고 있다. 일반적으로 유량을 추정하기 위해서 특정 단면에서의 수위를 측정하여 이를 수위-유량 관계곡선을 통해서 유량으로 환산하고, 수위-유량 관계를 측정한 후 이를 회귀분석 방법으로 내삽 및 외삽을 실시하여 유량을 측정하게 된다. 그러나 수위-유량 관계곡선에서 저수위와 고수위를 하나의 곡선식으로 하게 되는 경우 정도가 낮아지게 되므로 많은 경우에 있어서 저수위, 고수위를 각각의 곡선으로 구하여 사용하고 있다. 문제는 이러한 경우 정량적으로 변곡점을 구하기보다는 경험적으로 저수위와 고수위를 구분하고 있으며, 수위-유량관계를 회귀식에 의해서 추정하게 되므로 이에 대한 불확실성이 발생하게 된다. 따라서 본 연구에서는 불확실성을 정량화시키기 위한 방법으로 Bayesian MCMC 기법을 활용하며 수위-유량 관계곡선식의 매개변수들의 사후분포를 추정하여 매개변수의 최적화 및 불확실성을 평가하였다. 앞서 언급되었듯이 저수위 및 고수위로 분리하여 수위-유량 곡선식을 도출하고 있으나 저수위 및 고수위를 분리하는 기준이 경험적이기 때문에 신뢰성이 저해되는 문제점이 발생한다. 본 연구에서는 수위-유량 곡선식의 매개변수들을 최적화 하는 동시에 Poisson 분포 기반의 변동점 분석이 연동되어 저수위 및 고수위를 분리할 수 있는 Bayesian 기반 통합 수위-유량 곡선 해석 방법을 개발하고자 한다.

• Journal of Korea Water Resources Association
• /
• v.49 no.5
• /
• pp.399-410
• /
• 2016

In case of the application of Hazen-Williams C for design, operation or maintenance of water supply system, field situations always should be reflected on the factors. In this study, the relationships between C factors and influencing factors are analyzed using statistical techniques with 174 measured C factor data collected in periodic inspection for safety diagnosis in multi-regional water supply systems. To analyze their relationships, cross analysis, one-way ANOVA, correlation analysis were conducted. Analysis results showed that C factors had high correlations with both of elapsed year and pipe diameter and were relatively highly affected by coating material among influencing factors with the categorical type. On the other hand, elapsed year, pipe diameter and water type were meaningful influencing factors according to the results of multiple regression analysis. The Cluster analysis revealed that C factors had a tendency of being fundamentally classified on the basis of the elapsed year of about 20 years and the pipe diameter of 1500mm. Although C factors were generally greatly affected by elapsed year, size of pipe diameter relatively had an large influence on values of them in case of large diameter pipes. Lastly, It can be suggested that C factor estimation formulas using multiple regression analysis and clustering analysis in this study, can be applied as decision standards of C factor in multi-regional water supply systems.

• The Journal of the Korea Contents Association
• /
• v.12 no.3
• /
• pp.44-54
• /
• 2012

In wireless sensor networks, a moving object tracking scheme is one of core technologies for real applications such as environment monitering and enemy moving tracking in military areas. However, no works have been carried out on processing the failure of object tracking in sparse sensor networks with holes. Therefore, the energy consumption in the existing schemes significantly increases due to plenty of failures of moving object tracking. To overcome this problem, we propose a novel moving object tracking scheme based on polynomial regression prediction in sparse sensor networks. The proposed scheme activates the minimum sensor nodes by predicting the trajectory of an object based on polynomial regression analysis. Moreover, in the case of the failure of moving object tracking, it just activates only the boundary nodes of a hole for failure recovery. By doing so, the proposed scheme reduces the energy consumption and ensures the high accuracy for object tracking in the sensor network with holes. To show the superiority of our proposed scheme, we compare it with the existing scheme. Our experimental results show that our proposed scheme reduces about 47% energy consumption for object tracking over the existing scheme and achieves about 91% accuracy of object tracking even in sensor networks with holes.

• Korean Journal of Construction Engineering and Management
• /
• v.13 no.2
• /
• pp.91-101
• /
• 2012

In planning phase of Public Multi-Housing Projects, it is required to develop the methodology and criteria for fair cost prediction with influencing power from planning phase to occupancy phase. Many studies still have focused on the prediction of cost by multiple regression. However, there is no logical explanation about the influence of nonmetric variables for the prediction of cost in planning phase. Accordingly, this research pursues a cost prediction model including nonmetric variables for use in planning phase. There are 3 steps of this research : 1) Finding the factors influencing construction cost and assigning variables for a multiple regression. 2) Conducting a dummy regression analysis with nonmetric variables and model validation by comparing actual cost data. 3) Developing the ratio of RC structure cost to wall structure cost by using cost predection model. The results could establish cost prediction process including the influence of nonmetric variables and the ratio of RC structure cost to wall structure cost.