Bae J.K.,Dongyang University |
Kim J.,Sogang University
Expert Systems with Applications | Year: 2011
The knowledge-based artificial neural network (KBANN) is composed of phases involving the expression of domain knowledge, the abstraction of domain knowledge at neural networks, the training of neural networks, and finally, the extraction of rules from trained neural networks. The KBANN attempts to open up the neural network black box and generates symbolic rules with (approximately) the same predictive power as the neural network itself. An advantage of using KBANN is that the neural network considers the contribution of the inputs towards classification as a group, while rule-based algorithms like C5.0 measure the individual contribution of the inputs one at a time as the tree is grown. The knowledge consolidation model (KCM) combines the rules extracted using KBANN (NeuroRule), frequency matrix (which is similar to the Naïve Bayesian technique), and C5.0 algorithm. The KCM can effectively integrate multiple rule sets into one centralized knowledge base. The cumulative rules from other single models can improve overall performance as it can reduce error-term and increase R-square. The key idea in the KCM is to combine a number of classifiers such that the resulting combined system achieves higher classification accuracy and efficiency than the original single classifiers. The aim of KCM is to design a composite system that outperforms any individual classifier by pooling together the decisions of all classifiers. Another advantage of KCM is that it does not need the memory space to store the dataset as only extracted knowledge is necessary in build this integrated model. It can also reduce the costs from storage allocation, memory, and time schedule. In order to verify the feasibility and effectiveness of KCM, personal credit rating dataset provided by a local bank in Seoul, Republic of Korea is used in this study. The results from the tests show that the performance of KCM is superior to that of the other single models such as multiple discriminant analysis, logistic regression, frequency matrix, neural networks, decision trees, and NeuroRule. Moreover, our model is superior to a previous algorithm for the extraction of rules from general neural networks. © 2010 Elsevier Ltd. All rights reserved.
Bae J.K.,Dongyang University |
Kim J.,Sogang University
Expert Systems with Applications | Year: 2011
Many enterprises have been devoting a significant portion of their budget to product development in order to distinguish their products from those of their competitors and to make them better fit the needs and wants of customers. Hence, businesses should develop product designing that could satisfy the customers' requirements since this will increase the enterprise's competitiveness and it is an essential criterion to earning higher loyalties and profits. This paper investigates the following research issues in the development of new digital camera products: (1) What exactly are the customers' "needs" and "wants" for digital camera products? (2) What features is more importance than others? (3) Can product design and planning for product lines/product collection be integrated with the knowledge of customers? (4) How can the rules help us to make a strategy during we design new digital camera? To investigate these research issues, the Apriori and C5.0 algorithms are methodologies of association rules and decision trees for data mining, which is implemented to mine customer's needs. Knowledge extracted from data mining results is illustrated as knowledge patterns and rules on a product map in order to propose possible suggestions and solutions for product design and marketing. © 2011 Published by Elsevier Ltd.
Kim S.,Dongyang University |
Shiri J.,Islamic Azad University at Tabriz |
Kisi O.,Canik Basari University
Water Resources Management | Year: 2012
The purpose of this study was to develop and apply the neural networks models to estimate daily pan evaporation (PE) for different climatic zones such as temperate and arid climatic zones, Republic of Korea and Iran. Three kinds of the neural networks models, namely multilayer perceptron-neural networks model (MLP-NNM), generalized regression neural networks model (GRNNM), and support vector machine-neural networks model (SVM-NNM), were used to estimate daily PE. The available climatic variables, consisted of mean air temperature (T mean), mean wind speed (U mean), sunshine duration (SD), mean relative humidity (RH mean), and extraterrestrial radiation (R a) were used to estimate daily PE using the various input combinations of climate variables. The measurements for the period of January 1985-December 1990 (Republic of Korea) and January 2002-December 2008 (Iran) were used for training and testing the employed neural networks models. The results obtained by SVM-NNM indicated that it performs better than MLP-NNM and GRNNM for estimating daily PE. A comparison was also made among the employed models, which demonstrated the superiority of MLP-NNM, GRNNM, and SVM-NNM over Linacre model and multiple linear regression model (MLRM). © 2012 Springer Science+Business Media B.V.
Bae J.K.,Dongyang University
Journal of Convergence Information Technology | Year: 2010
In this study, performance of classification techniques is compared in order to predict dividend policy decisions. We first analyzed the feasibility of all available companies listed in the Korea Exchange (KRX) market as dividend data sets by using classification techniques. Then we developed a prediction model based on support vector machines (SVM). We compare the classification accuracy performance between our SVM model and artificial intelligence techniques, and suggest a better dividend policy forecasting model to help a chief executive officer (CEO) or a board of directors (BOD) make better decision in a corporate dividend policy. The experiments demonstrate that the SVM model always outperforms other models in the performance of dividend policy forecasting, and hence we can predict future dividend policy more correctly than any other models. This enhancement in predictability of future dividend policy can significantly contribute to the correct valuation of a company, and hence those people from investors to financial managers to any decision makers of a company can make use of the SVM model for the better financing and investing decision making which can lead to higher profits and firm values eventually. Moreover, this is particularly important for people who want to obtain a high level of accuracy in advanced areas such as financial decision makings.
Cho G.-S.,Dongyang University
Computers and Security | Year: 2013
In this paper, we present a computer forensic method for detecting timestamp forgeries in the Windows NTFS file system. It is difficult to know precisely that the timestamps have been changed by only examining the timestamps of the file itself. If we can find the past timestamps before any changes to the file are made, this can act as evidence of file time forgery. The log records operate on files and leave large amounts of information in the $LogFile that can be used to reconstruct operations on the files and also used as forensic evidence. Log record with 0x07/0x07 opcode in the data part of Redo/Undo attribute has timestamps which contain past-and-present timestamps. The past-and-present time-stamps can be decisive evidence to indicate timestamp forgery, as they contain when and how the timestamps were changed. We used file time change tools that can easily be found on Internet sites. The patterns of the timestamp change created by the tools are different compared to those of normal file operations. Seven file operations have ten timestamp change patterns in total by features of timestamp changes in the $STANDARD-INFORMATION attribute and the $FILE-NAME attribute. We made rule sets for detecting timestamp forgery based on using difference comparison between changes in timestamp patterns by the file time change tool and normal file operations. We apply the forensic rule sets for ".txt", ".docx" and ".pdf" file types, and we show the effectiveness and validity of the proposed method. The importance of this research lies in the fact that we can find the past time in $LogFile, which gives decisive evidence of timestamp forgery. This makes the timestamp active evidence as opposed to simply being passive evidence. © 2012 Elsevier Ltd. All rights reserved.
Kim S.,Dongyang University
Disaster Advances | Year: 2011
The goal of this research is to develop and apply the integrational operation method (IOM) for modeling the relationship of the pan evaporation (PE) and the alfalfa reference evapotranspiration (ETr). Since the observed data of the alfalfa ETr using lysimeter have not been measured for a long time, the Penman-Monteith (PM) method is used to estimate the observed alfalfa ETr The IOM consists of the combination/ application of the stochastic and neural networks models respectively. The stochastic model of Periodic Auto Regressive Moving Average (PARMA) is applied to generate the training dataset for the monthly PE and the alfalfa ET r and the neural networks models are applied to calculate the observed test dataset reasonably. Among the six training patterns, 1,000/PARMA(1,1) /GRNNM-GA training pattern is used which can evaluate the suggested climatic variables very well and construct the reliable data for the monthly PE and the alfalfa ETr. Uncertainty analysis is also used to eliminate the climatic variables of input nodes from the 1,000/PARMA(1, 1)ZGRNNM-GA training pattern. The sensitive and insensitive climatic variables are chosen from the uncertainty analysis of the input nodes.
Bae S.-K.,Dongyang University
Journal of Information Processing Systems | Year: 2014
Various Time Synchronization protocols for a Wireless Sensor Network (WSN) have been developed since time synchronization is important in many timecritical WSN applications. Aside from synchronization accuracy, energy constraint should also be considered seriously for time synchronization protocols in WSNs, which typically have limited power environments. This paper performs analysis of prominent WSN time synchronization protocols in terms of power consumption and test by simulation. In the analysis and simulation tests, each protocol shows different performance in terms of power consumption. This result is helpful in choosing or developing an appropriate time synchronization protocol that meets the requirements of synchronization accuracy and power consumption (or network lifetime) for a specific WSN application. © 2014 KIPS.
Park W.-D.,Dongyang University |
Tanioka K.,Tokyo Denki University
Applied Physics Letters | Year: 2014
Amorphous selenium (a-Se) high-gain avalanche rushing amorphous photoconductor (HARP) film has been used for highly sensitive imaging devices. To improve the spectral response of a-Se HARP photoconductive film at a long wavelength, the tellurium (Te) doping effect in an 8-μm-thick a-Se HARP film was investigated. The thickness of the Te-doped a-Se layer in the 8-μm-thick a-Se HARP films was varied from 60 to 120 nm. The signal current increases significantly due to the avalanche multiplication when the target voltage is increased over the threshold voltage. In the 8-μm-thick a-Se HARP film with a Te-doped layer, the spectral response at a long wavelength was improved in comparison with the a-Se HARP film without a Te-doped layer. In addition, the increase of the lag in the 8-μm-thick a-Se HARP target with a Te-doped layer of 120 nm is caused by the photoconductive lag due to the electrons trapped in the Te-doped layer. Based on the current-voltage characteristics, spectral response, and lag characteristics of the 8-μm-thick a-Se HARP targets, the Te-doped layer thickness of 90 nm is suitable for the 8-μm-thick a-Se HARP film. © 2014 AIP Publishing LLC.
Park W.-D.,Dongyang University
Transactions on Electrical and Electronic Materials | Year: 2012
CdS thin film was prepared on glass substrate by chemical bath deposition in an alkaline solution. The optical properties of CdS thin film were investigated using spectroscopic ellipsometry. The real (ε 1) and imaginary (ε 2) parts of the complex dielectric function ε(E)=ε 1(E) + iε 2(E), the refractive index n(E), and the extinction coefficient R(E) of CdS thin film were obtained from spectroscopic ellipsometry. The normal-incidence reflectivity R(E) and absorption coefficient α(E) of CdS thin film were obtained using the refractive index and extinction coefficient. The critical points E 0 and E 1 of CdS thin film were shown in spectra of the dielectric function and optical constants of refractive index, extinction coefficient, normal-incidence reflectivity, and absorption coefficient. The dispersion of refractive index was analyzed by the Wemple-DiDomenico single-oscillator model. © 2012 KIEEME. All rights reserved.
Do J.H.,Dongyang University
Molecules and cells | Year: 2014
The exact causes of cell death in Parkinson's disease (PD) remain unknown despite extensive studies on PD.The identification of signaling and metabolic pathways involved in PD might provide insight into the molecular mechanisms underlying PD. The neurotoxin 1-methyl-4-phenylpyridinium (MPP(+)) induces cellular changes characteristic of PD, and MPP(+)-based models have been extensively used for PD studies. In this study, pathways that were significantly perturbed in MPP(+)-treated human neuroblastoma SH-EP cells were identified from genome-wide gene expression data for five time points (1.5, 3, 9, 12, and 24 h) after treatment. The mitogen-activated protein kinase (MAPK) signaling pathway and endoplasmic reticulum (ER) protein processing pathway showed significant perturbation at all time points. Perturbation of each of these pathways resulted in the common outcome of upregulation of DNA-damage-inducible transcript 3 (DDIT3). Genes involved in ER protein processing pathway included ubiquitin ligase complex genes and ER-associated degradation (ERAD)-related genes. Additionally, overexpression of DDIT3 might induce oxidative stress via glutathione depletion as a result of overexpression of CHAC1. This study suggests that upregulation of DDIT3 caused by perturbation of the MAPK signaling pathway and ER protein processing pathway might play a key role in MPP(+)-induced neuronal cell death. Moreover, the toxicity signal of MPP(+) resulting from mitochondrial dysfunction through inhibition of complex I of the electron transport chain might feed back to the mitochondria via ER stress. This positive feedback could contribute to amplification of the death signal induced by MPP(+).