• Journal of the Korea Society of Computer and Information
• /
• v.24 no.12
• /
• pp.1-7
• /
• 2019

With the in-depth development of machine learning, the deep learning method has made great progress, especially with the Convolution Neural Network(CNN). Compared with traditional text sentiment classification methods, deep learning based CNNs have made great progress in text classification and processing of complex multi-label and multi-classification experiments. However, there are also problems with the neural network for text sentiment classification. In this paper, we propose a fusion model based on Long-Short Term Memory networks(LSTM) and CNN deep learning methods, and applied to multi-category news datasets, and achieved good results. Experiments show that the fusion model based on deep learning has greatly improved the precision and accuracy of text sentiment classification. This method will become an important way to optimize the model and improve the performance of the model.

• The Journal of the Korea Contents Association
• /
• v.10 no.4
• /
• pp.19-27
• /
• 2010

In recent years, high-throughput genome-wide RNA interference screening is emerging as an essential tool to biologists in understanding complex cellular processes. The manual analysis of the large number of images produced in each study spends much time and the labor. Hence, automatic cellular image analysis becomes an urgent need, where segmentation is the first and one of the most important steps. However, those factors such as the region overlapping, a variety of shapes, and non-uniform local characteristics of cellular images become obstacles to efficient cell segmentation. To avoid the problem, a new watershed-based cell segmentation algorithm using a localized segmentation method and a feature vector is proposed in this paper. Localized approach in segmentation resolves the problems caused by a variety of shapes and non-uniform characteristics. In addition, the poor performance of segmentation in overlapped regions can be improved by taking advantage of a feature vector whose component features complement each other. Simulation results show that the proposed method improves the segmentation performance compared to the method in Cellprofiler.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.33 no.12A
• /
• pp.1156-1164
• /
• 2008

In this paper, we propose a bi-directional coding method (BTC) using time reversal mirror for effective data exchange between mobile stations and basestations. In conventional relay systems, while one node transmits data, other nodes should keep silent to avoid interference. Thus, to interchange data between basestations and mobile stations, it needs four stages. However, using the proposed BTC method, a relay station transmits a probe source before data transmission, and using this information, mobile stations and basestations transmit data to the relay station simultaneously. Therefore, since it exchanges data in only two steps, the system achieves improved system capacity. In addition, owing to no need for complex calculation, the relay system can be implemented with low complexity. The simulation results demonstrate that the proposed method achieves enhanced system capacity.

• Journal of the Korea Safety Management & Science
• /
• v.9 no.6
• /
• pp.39-46
• /
• 2007

In this paper we discuss the problems related with the existing electrical safety system that is influenced by all the socio-technical factors including the electrical safety law, government organizations, and their practicing rules and regulations. Sometimes intangible practices also affect the safety issues. It is our intention to incorporate all those factors effectively and systematically in improving the electrical safety system. Since, the problem in nature is so complex, an approach based on the systems engineering (SE) method could be a good candidate. Specifically, as a first step of the SE process we performed the requirements analysis of all the factors mentioned above and some recommendations from the experts in the field. In the work environment utilizing a computer-aided SE tool, the requirements are verified and managed efficiently although the data base is complicated in general in tracking the changes. Next, by analyzing the verified requirements, a functional architecture is developed to achieve an improved electrical safety system. We also did the reverse engineering of the existing electrical safety system to explore potential problems there. The result can also help improve the existing one. Overall, a systematic approach to constructing electrical safety system based on SE is proposed for the first time in this paper. Note that the method described here can be adopted in the study of other safety systems.

• Journal of Korea Water Resources Association
• /
• v.42 no.11
• /
• pp.989-1004
• /
• 2009

The objective of this study is to examine validation of Grid Reconstruction Method, which is developed to simulate drying/wetting in complex natural rivers with wetting and drying domain areas. To verify application of the developed model, the model was applied to natural rivers with wetting and drying domain areas such as Han river and Nakdong river. The simulation results have shown good agreements with observed data and the results for the developed model were more accurate and improved stability of numerical computation than those of RMA-2 model. If the analysis of contaminant advection-diffusion and sediment transport are performed with the study results, the results can be effectively applied to river flow analysis and ecological hydraulics.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.6 no.2
• /
• pp.81-89
• /
• 1996

A design method of fuzzy modeling is presented for the model identification of route choice of traffic problems.The proposed fuzzy modeling implements system structure and parameter identification in the eficient form of""IF..., THEN-.."", using the theories of optimization theory, linguistic fuzzy implication rules. Three kinds ofmethod for fuzzy modeling presented in this paper include simplified inference (type I), linear inference (type 21,and proposed modified-linear inference (type 3). The fuzzy inference method are utilized to develop the routechoice model in terms of accurate estimation and precise description of human travel behavior. In order to identifypremise structure and parameter of fuzzy implication rules, improved complex method is used and the least squaremethod is utilized for the identification of optimum consequence parameters. Data for route choice of trafficproblems are used to evaluate the performance of the proposed fuzzy modeling. The results show that the proposedmethod can produce the fuzzy model with higher accuracy than previous other studies -BL(binary logic) model,B(production system) model, FL(fuzzy logic) model, NN(neura1 network) model, and FNNs (fuzzy-neuralnetworks) model -.fuzzy-neural networks) model -.

• Journal of the Korea Society for Simulation
• /
• v.24 no.4
• /
• pp.11-21
• /
• 2015

If the cloud computing environment is not sufficiently provide the required resources due to the number of virtual server to process the request, may cause a problem that the load applied to the specific server. Migration administrator receive the resources of each physical server for improving the efficiency of the virtual server that exists in the physical servers, and determines the migration destination based on the simulation results. But, there is more overhead predicting the future resource consumption of all the physical server to decide the migration destination through the simulation process in large and complex cloud computing environments. To solve this problem, we propose an improved prediction method with the simulation-based approach. The proposed method is a fuzzy-logic based priority model for VM migration. We design a proposed model with the DEVS formalism. And we also measure and compare a performance and migration count with existing simulation-based migration method. FPLM shows high utilization.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.42 no.5 s.305
• /
• pp.145-156
• /
• 2005

In application for acoustic echo cancellation and adaptive equalization, input signal is highly correlated and the long length of adaptive filter is needed. Affine projection algorithms, in these applications, can produce a good convergence performance. However, they have a drawback that is a complex hardware implementation. In this paper, we propose a new subband affine projection algorithm with improved convergence and reduced computational complexity. In addition, we suggest a good approach to implement the proposed method. In this method by applying polyphase decomposition, noble identity and critical decimation to the anne projection algorithm the number of input vectors for decorrelation can be reduced. The weight-updating formula of the proposed method is derived as a simple form that compared with the NLMS(normalized least mean square) algorithm by the reduced projection order The efficiency of the proposed algorithm for a colored input signal was evaluated by using computer simulations.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.22 no.1
• /
• pp.17-25
• /
• 2018

Recently, with the development of ICT technology and the rapid spread of smart devices, a huge amount of information is being generated. The recommendation system has helped the informant to judge the information from the information overload, and it has become a solution for the information provider to increase the profit of the company and the publicity effect of the company. Recommendation systems can be implemented in various approaches, but social information is presented as a way to improve performance. However, no research has been done to utilize trust cluster information among users in the recommendation system. In this paper, we propose a method to improve the performance of the recommendation system by using the influence between the intra-cluster objects and the information between the trustor-trustee in the cluster generated in the online review. Experiments using the proposed method and real data have confirmed that the prediction accuracy is improved than the existing methods.

• Proceedings of the KIEE Conference
• /
• 1998.07c
• /
• pp.918-922
• /
• 1998

With the development of industry, the qualitical advancement of power is needed. Since it is placed in the end step of power system, the fault at the distribution system causes some users blackout directly. So if the fault occurs, quick restoration is very important subject and, for the reason, induction of the distribution automation system is now being progressed briskly. For the quick restoration of the faulted distribution system, the load shedding of the blackout-area must be followed, and the other problems like the shedded load, faulted voltage and the rest may cause other accident. Accordingly load shedding must be based on the precise calculation technique during the distribution system load flow(dist flow) calculation. In these days because of its superior convergence characteristic the Newton-Raphson method is most widely used. The number of buses in the distribution system amounts to thousands, and if the fault occurs at the distribution system, the speed for the dist flow calculation is to be improved to apply to the On-Line system. However, Newton-Raphson method takes much time relatively because it must calculate the Jacobian matrix and inverse matrix at every iteration, and in the case of huge load, the equation is hard to converge. In this thesis. matrix equation is used to make algebraical expression and then to solve load flow equation and to modify above defects. Then the complex matrix is divided into real part and imaginary part to keep sparcity. As a result time needed for calculation diminished. Application of mentioned algorithm to 302 bus, 700 bus, 1004 bus system led to almost identical result got by Newton-Raphson method and showed constant convergence characteristic. The effect of time reduction showed 88.2%, 86.4%, 85.1% at each case of 302 bus, 700 bus system 86.4%, and 1004 bus system.