• Journal of the Korean Data and Information Science Society
• /
• v.19 no.4
• /
• pp.1219-1231
• /
• 2008

The main object of this paper is to develop a leave-one-out(LOO) bound of all pairwise comparison error correcting output codes (APC-ECOC). To avoid using classifiers whose corresponding target values are 0 in APC-ECOC and requiring pilot estimates we developed a bound based on mean misclassification probability(MMP). It can be used to tune kernel hyperparameters. Our empirical experiment using kernel mean squared estimate(KMSE) as the binary classifier indicates that the bound leads to good estimates of kernel hyperparameters.

• Proceedings of the IEEK Conference
• /
• 1998.10a
• /
• pp.1277-1280
• /
• 1998

We propose a structural learning method of MLP classifiers for a given application using PfSGA (parameter-free species genetic algorithm), which is a combining of species genetic algorithm(SGA) and parameter-free genetic algorithm(PfGA). experimental results show that PfSGA can reduce the learing time of SGA and has no influence of parameter values on structural learning. And we also convince that PfSGA is more efficient than the other methods in the aspect of misclassification ratio, learning rate, and complexity of MLP structure.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.12 no.1
• /
• pp.1-6
• /
• 1975

The boundary has been one of the most important information in radiographic images and the degrees of difficulty involved varies greatly with the quality of the picture. These Buantifications are the means to diagnoses. The purpose of this paper is to quantify intensity variation and the threshold decision which is based on statistical principles and is developed to detect limits in liver scintigrams the entire picture is devide4 into 64 small regions. The kurtosis and variances for each smal region are used as indications to select the histograms the thresholds are computed according to the method o(maximum likelihood which minimizes the probability o( misclassification. Therefore Ive have demonstrated the applicability of the boundary detection and proved good agreement with human recognition, and we can use it for the diagnosis data of liver disease.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.40 no.3
• /
• pp.194-202
• /
• 2003

To convert printed images into electronic ones automatically, it requires region analysis of document images and character recognition. In these, regional analysis segments document image into detailed regions and classifies thee regions into the types of text, picture, table and so on. But it is difficult to classify the text and the picture exactly, because the size, density and complexity of pixel distribution of some of these are similar. Thu, misclassification in region analysis is the main reason that makes automatic conversion difficult. In this paper, we propose region analysis method that segments document image into text and picture regions. The proposed method solves the referred problems using one dimensional median filter based method in text and picture classification. And the misclassification problems of boldface texts and picture regions like graphs or tables, caused by using median filtering, are solved by using of skin peeling filter and maximal text length. The performance, therefore, is better than previous methods containing commercial softwares.

• Journal of Intelligence and Information Systems
• /
• v.9 no.1
• /
• pp.71-89
• /
• 2003

Although many studies demonstrate that one technique outperforms the others for a given data set, there is often no way to tell a priori which of these techniques will be most effective in the classification problems. Alternatively, it has been suggested that a better approach to classification problem might be to integrate several different forecasting techniques. This study proposes the linearly combining methodology of different classification techniques. The methodology is developed to find the optimal combining weight and compute the weighted-average of different techniques' outputs. The proposed methodology is represented as the form of mixed integer programming. The objective function of proposed combining methodology is to minimize total misclassification cost which is the weighted-sum of two types of misclassification. To simplify the problem solving process, cutoff value is fixed and threshold function is removed. The form of mixed integer programming is solved with the branch and bound methods. The result showed that proposed methodology classified more accurately than any of techniques individually did. It is confirmed that Proposed methodology Predicts significantly better than individual techniques and the other combining methods.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.22 no.4
• /
• pp.405-413
• /
• 2012

Low misclassification performance is significant with high classification accuracy for a reliable diagnosis of ECG signals, and diagnosing abnormal state as normal state can especially raises a deadly problem to a person in ECG test. In this paper, we propose detection and classification method of abnormal rhythm by rule-based rhythm classification reflecting clinical criteria for disease. Rule-based classification classifies rhythm types using rule-base for feature of rhythm section, and rule-base deduces decision results corresponding to professional materials of clinical and internal fields. Experimental results for the MIT-BIH arrhythmia database show that the applicability of proposed method is confirmed to classify rhythm types for normal sinus, paced, and various abnormal rhythms, especially without misclassification in detection aspect of abnormal rhythm.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2002.05a
• /
• pp.1044-1051
• /
• 2002

This paper proposes a two-phase mathematical programming approach by considering classification gap to solve the proposed credit scoring problem so as to complement any theoretical shortcomings. Specifically, by using the linear programming (LP) approach, phase 1 is to make the associated decisions such as issuing grant of credit or denial of credit to applicants. or to seek any additional information before making the final decision. Phase 2 is to find a cut-off value, which minimizes any misclassification penalty (cost) to be incurred due to granting credit to 'bad' loan applicant or denying credit to 'good' loan applicant by using the mixed-integer programming (MIP) approach. This approach is expected to and appropriate classification scores and a cut-off value with respect to deviation and misclassification cost, respectively. Statistical discriminant analysis methods have been commonly considered to deal with classification problems for credit scoring. In recent years, much theoretical research has focused on the application of mathematical programming techniques to the discriminant problems. It has been reported that mathematical programming techniques could outperform statistical discriminant techniques in some applications, while mathematical programming techniques may suffer from some theoretical shortcomings. The performance of the proposed two-phase approach is evaluated in this paper with line data and loan applicants data, by comparing with three other approaches including Fisher's linear discriminant function, logistic regression and some other existing mathematical programming approaches, which are considered as the performance benchmarks. The evaluation results show that the proposed two-phase mathematical programming approach outperforms the aforementioned statistical approaches. In some cases, two-phase mathematical programming approach marginally outperforms both the statistical approaches and the other existing mathematical programming approaches.

• Journal of the Korean Data and Information Science Society
• /
• v.22 no.6
• /
• pp.1017-1028
• /
• 2011

When it is not easy to decide the credit scoring for some loan applicants, credit evaluation is postponded and reserve to ask a specialist for further evaluation of undecided applicants. This undecided inference is one of problems that happen to most statistical models including the biostatistics and sportal statistics as well as credit evaluation area. In this work, the undecided inference is regarded as a missing data mechanism under the assumption of MNAR, and use the bivariate probit model which is one of sample selection models. Two undecided inference methods are proposed: one is to make use of characteristic variables to represent the state for decided applicants, and the other is that more accurate and additional informations are collected and apply these new variables. With an illustrated example, misclassification error rates for undecided and overall applicants are obtainded and compared according to various characteristic variables, undecided intervals, and thresholds. It is found that misclassification error rates could be reduced when the undecided interval is increased and more accurate information is put to model, since more accurate situation of decided applications are reflected in the bivariate probit model.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.32 no.5
• /
• pp.997-1008
• /
• 2022

Today, AI(Artificial Intelligence) technology is being extensively researched in various fields, including the field of malware detection. To introduce AI systems into roles that protect important decisions and resources, it must be a reliable AI model. AI model that dependent on training dataset should be verified to be robust against new attacks. Rather than generating new malware detection, attackers find malware detection that succeed in attacking by mass-producing strains of previously detected malware detection. Most of the attacks, such as adversarial attacks, that lead to misclassification of AI models, are made by slightly modifying past attacks. Robust models that can be defended against these variants is needed, and the Robustness level of the model cannot be evaluated with accuracy and recall, which are widely used as AI evaluation indicators. In this paper, we experiment a framework to evaluate robustness level by generating an adversarial sample based on one of the adversarial attacks, C&W attack, and to improve robustness level through adversarial training. Through experiments based on malware dataset in this study, the limitations and possibilities of the proposed method in the field of malware detection were confirmed.

• International Journal of Computer Science & Network Security
• /
• v.24 no.5
• /
• pp.11-20
• /
• 2024

While artificial neural networks are adept at identifying patterns, they can struggle to distinguish between actual correlations and false associations between extracted facial features and criminal behavior within the training data. These associations may not indicate causal connections. Socioeconomic factors, ethnicity, or even chance occurrences in the data can influence both facial features and criminal activity. Consequently, the artificial neural network might identify linked features without understanding the underlying cause. This raises concerns about incorrect linkages and potential misclassification of individuals based on features unrelated to criminal tendencies. To address this challenge, we propose a novel region-based training approach for artificial neural networks focused on criminal propensity detection. Instead of solely relying on overall facial recognition, the network would systematically analyze each facial feature in isolation. This fine-grained approach would enable the network to identify which specific features hold the strongest correlations with criminal activity within the training data. By focusing on these key features, the network can be optimized for more accurate and reliable criminal propensity prediction. This study examines the effectiveness of various algorithms for criminal propensity classification. We evaluate YOLO versions YOLOv5 and YOLOv8 alongside VGG-16. Our findings indicate that YOLO achieved the highest accuracy 0.93 in classifying criminal and non-criminal facial features. While these results are promising, we acknowledge the need for further research on bias and misclassification in criminal justice applications