• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.22 no.3
• /
• pp.20-37
• /
• 2023

Although traffic signal installations are continuously expanding, the effect of preventing traffic accidents remains unverified. Totally, 7,045 traffic accident data (such as signal violations) registered with TCS were manually searched for a 7-year period from 2013 to 2019 for 1,602 traffic signals in Daejeon Metropolitan City. The top 20 traffic accident intersections were identified, the traffic accident investigation records and field maps were viewed to compare the driving direction and signal phase of the violated vehicle, and the cause of the traffic accident was divided into insufficient signal operation design (operation) and driver negligence (intentional). Results of the analysis revealed that 75% of traffic accidents occurred in thru-left-turn traffic signals and overlap; moreover, extending the yellow time or operating all red signals due to countermeasures against traffic accidents occurring in yellow signals resulted in reduced traffic accidents. Data indicated that Permissive Left Turn requires improvement with the signal operation. In addition, since The Korean National Police Agency is not computerized for traffic accident sites and signal-related data, the lack of manpower necessitates improvement and utilization of TCS when establishing traffic accident prevention measures. It is believed that it will contribute to signal operation by analyzing vast amounts of data collected in the field and presenting improvement measures.

• Journal of Intelligence and Information Systems
• /
• v.28 no.3
• /
• pp.161-183
• /
• 2022

Opinion polls have become a powerful means for election campaigns and one of the most important subjects in the media in that they predict the actual election results and influence people's voting behavior. However, the more active the polls, the more often they fail to properly reflect the voters' minds in measuring the effectiveness of election campaigns, such as repeatedly conducting polls on the likelihood of winning or support rather than verifying the pledges and policies of candidates. Even if the poor predictions of the election results of the polls have undermined the authority of the press, people cannot easily let go of their interest in polls because there is no clear alternative to answer the instinctive question of which candidate will ultimately win. In this regard, we attempt to retrospectively grasp public opinion on the 20th presidential election by applying the 'YouTube Analysis' function of Sometrend, which provides an environment for discovering insights through online big data. Through this study, it is confirmed that a result close to the actual public opinion (or opinion poll results) can be easily derived with simple YouTube data results, and a high-performance public opinion prediction model can be built.

• ETRI Journal
• /
• v.42 no.1
• /
• pp.78-89
• /
• 2020

Human activity recognition (HAR) has become effective as a computer vision tool for video surveillance systems. In this paper, a novel biometric system that can detect human activities in 3D space is proposed. In order to implement HAR, joint angles obtained using an RGB-depth sensor are used as features. Because HAR is operated in the time domain, angle information is stored using the sliding kernel method. Haar-wavelet transform (HWT) is applied to preserve the information of the features before reducing the data dimension. Dimension reduction using an averaging algorithm is also applied to decrease the computational cost, which provides faster performance while maintaining high accuracy. Before the classification, a proposed thresholding method with inverse HWT is conducted to extract the final feature set. Finally, the K-nearest neighbor (k-NN) algorithm is used to recognize the activity with respect to the given data. The method compares favorably with the results using other machine learning algorithms.

• Proceedings of the KSRS Conference
• /
• 2005.10a
• /
• pp.337-340
• /
• 2005

Context-awareness techniques in ubiquitous computing environment provide various services to users who need to get information via the analysis of collected information from sensors in a spatial area. Context-awareness has been increased in ubiquitous computing and is applied to many different applications such as disaster management system, intelligent robot system, transportation management system, shopping management system, and digital home service. Many researches have recently focused on services that provide the appropriate information, which are collected from Internet by different kinds of sensors, to users according to context of their surrounding environment. In this paper, we propose an abstraction model to manage the large-scale contextual information and their metadata which are collected from different kinds of in-situ sensors in a spatial area and are presented them on the web. This model is composed of the modules expressing functional elements of sensors using sensorML(Sensor Model Language) based on XML language and the modules managing contextual information, which is transmitted from the sensors.

• Journal of Intelligence and Information Systems
• /
• v.5 no.1
• /
• pp.95-101
• /
• 1999

In this paper, a comparison between different methods to combine predictions from neural networks will be given. These methods are bagging, bumping, and balancing. Those are based on the analysis of the ensemble generalization error into an ambiguity term and a term incorporating generalization performances of individual networks. Neural Networks and AI machine learning models are prone to overfitting. A strategy to prevent a neural network from overfitting, is to stop training in early stage of the learning process. The complete data set is spilt up into a training set and a validation set. Training is stopped when the error on the validation set starts increasing. The stability of the networks is highly dependent on the division in training and validation set, and also on the random initial weights and the chosen minimization procedure. This causes early stopped networks to be rather unstable: a small change in the data or different initial conditions can produce large changes in the prediction. Therefore, it is advisable to apply the same procedure several times starting from different initial weights. This technique is often referred to as training ensembles of neural networks. In this paper, we presented a comparison of three statistical methods to prevent overfitting of neural network.

• Proceedings of the Korea Inteligent Information System Society Conference
• /
• 2000.04a
• /
• pp.235-241
• /
• 2000

This article suggests integrated neural network models for the interest rate forecasting using change point detection. The basic concept of proposed model is to obtain intervals divided by change point, to identify them as change-point groups, and to involve them in interest rate forecasting. the proposed models consist of three stages. The first stage is to detect successive change points in interest rate dataset. The second stage is to forecast change-point group with data mining classifiers. The final stage is to forecast the desired output with BPN. Based on this structure, we propose three integrated neural network models in terms of data mining classifier: (1) multivariate discriminant analysis (MDA)-supported neural network model, (2) case based reasoning (CBR)-supported neural network model and (3) backpropagation neural networks (BPN)-supported neural network model. Subsequently, we compare these models with a neural networks (BPN)-supported neural network model. Subsequently, we compare these models with a neural network model alone and, in addition, determine which of three classifiers (MDA, CBR and BPN) can perform better. This article is then to examine the predictability of integrated neural network models for interest rate forecasting using change-point detection.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.20 no.3
• /
• pp.59-64
• /
• 2020

In the Paris Climate Agreement, Korea submitted an ambitious goal of reducing the greenhouse gas emission forecast (BAU) by 37% by 2030. And as one of the countermeasures, a smart grid, an intelligent power grid, was presented. In order to apply the smart grid, EMS(Energy Management System) needs to be installed and operated in various fields, and the supply is delayed due to the lack of awareness of users and the limitations of system ROI. Therefore, recently, various data analysis and control technologies have been proposed to increase the efficiency of the installed EMS. In this study, we present a measurement control algorithm that analyzes and predicts big data collected by IoT using a SARIMA model to check and operate energy consumption of public sports facilities.

• Journal of Korea Port Economic Association
• /
• v.17 no.2
• /
• pp.167-182
• /
• 2001

The government of Busan Metropolitan City conducted the "Survey on Shipping and Port Logistics Companies in Busan" between May and December, 2000. This was the first comprehensive survey of Busan Port conducted from the perspective of regional & industrial economics. The objective of the survey was to find out the level of influence of the shipping and port logistics industry on the regional economy of Busan, and to obtain base data for use in establishing an actual promotion program. The survey acquired information about human and physical resources, management conditions and consciousness of the industry, In addition, the study analyzed the survey data. The results of the analysis showed a method of creating added value in support of the marketing activities of the companies, and indicated methods of achieving systematic and sustainable promotion. The government of Busan City shall develop an e-Logistics infrastructure which can deliver a synopsis and intelligent information to people and companies in the industry by the end of this year. The information system would be of great help for people who may not be familiar with Busan's port and shipping industry, particularly international shipping companies. This will result in an Increase of trade and exchange in the shipping and port logistics industry, resulting in the generation of increased added value within the near future.

• Transactions of the Korean Society for Noise and Vibration Engineering
• /
• v.22 no.5
• /
• pp.413-421
• /
• 2012

In industry, various rotating machinery such as pumps, gas turbines, compressors, electric motors, generators are being used as an important facility. Due to the industrial development, they make high performance(high-speed, high-pressure). As a result, we need more intelligent and reliable machine condition diagnosis techniques. Diagnosis technique using hidden Markov-model is proposed for an accurate and predictable condition diagnosis of various rotating machines and also has overcame the speed limitation of time/frequency method by using compensation of the rotational speed of rotor. In addition, existing artificial intelligence method needs defect state data for fault detection. hidden Markov model can overcome this limitation by using normal state data alone to detect fault of rotational machinery. Vibration analysis of step-up gearbox for wind turbine was applied to the study to ensure the robustness of diagnostic performance about compensation of the rotational speed. To assure the performance of normal state alone method, hidden Markov model was applied to experimental torque measuring gearbox in this study.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.8 no.9
• /
• pp.1313-1318
• /
• 2013

In order to actively respond to cyber attacks, not only the security systems such as IDS, IPS, and Firewalls, but also ESM, a system that detects cyber attacks by analyzing various log data, are preferably deployed. However, as the attacks be come more elaborate and advanced, existing signature-based detection methods start to face their limitations. In response to that, researches upon symptom detection technology based on attack modeling by employing big-data analysis technology are actively on-going. This symptom detection technology is effective when it can accurately extract features of attacks and manipulate them to successfully execute the attack modeling. We propose the ways to extract attack features which can play a role as the basis of the modeling and detect intelligent threats by carrying out scenario-based modeling.