• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.1
• /
• pp.27-32
• /
• 2019

Research that implements the classification process through Deep Learning algorithm, one of the outstanding human abilities, includes a unimodal model, a multi-modal model, and a multi-modal method using music videos. In this study, the results were better by suggesting a system to analyze each song's spectrum into short samples and vote for the results. Among Deep Learning algorithms, CNN showed superior performance in the category of music genre compared to RNN, and improved performance when CNN and RNN were applied together. The system of voting for each CNN result by Deep Learning a short sample of music showed better results than the previous model and the model with Softmax layer added to the model performed best. The need for the explosive growth of digital media and the automatic classification of music genres in numerous streaming services is increasing. Future research will need to reduce the proportion of undifferentiated songs and develop algorithms for the last category classification of undivided songs.

• Korean Journal of Remote Sensing
• /
• v.36 no.6_2
• /
• pp.1591-1604
• /
• 2020

This study is a study on classifying land cover by applying high-resolution satellite images to deep learning algorithms and verifying the performance of algorithms for each spatial object. For this, the Fully Convolutional Network-based algorithm was selected, and a dataset was constructed using Kompasat-3 satellite images, land cover maps, and forest maps. By applying the constructed data set to the algorithm, each optimal hyperparameter was calculated. Final classification was performed after hyperparameter optimization, and the overall accuracy of DeeplabV3+ was calculated the highest at 81.7%. However, when looking at the accuracy of each category, SegNet showed the best performance in roads and buildings, and U-Net showed the highest accuracy in hardwood trees and discussion items. In the case of Deeplab V3+, it performed better than the other two models in fields, facility cultivation, and grassland. Through the results, the limitations of applying one algorithm for land cover classification were confirmed, and if an appropriate algorithm for each spatial object is applied in the future, it is expected that high quality land cover classification results can be produced.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.31 no.6
• /
• pp.1227-1236
• /
• 2021

Adversarial attack, one of the attacks on deep learning classification model, is attack that add indistinguishable perturbations to input data and cause deep learning classification model to misclassify the input data. There are various adversarial attack algorithms. Accordingly, many studies have been conducted to detect adversarial attack but few studies have been conducted to classify what adversarial attack algorithms to generate adversarial input. if adversarial attacks can be classified, more robust deep learning classification model can be established by analyzing differences between attacks. In this paper, we proposed a model that detects and classifies adversarial attacks by constructing a random forest classification model with input features extracted from a target deep learning model. In feature extraction, feature is extracted from a output value of hidden layer based on class predicted by the target deep learning model. Through Experiments the model proposed has shown 3.02% accuracy on clean data, 0.80% accuracy on adversarial data higher than the result of pre-existing studies and classify new adversarial attack that was not classified in pre-existing studies.

• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.10
• /
• pp.437-446
• /
• 2022

Text classification task from Natural Language Processing (NLP) combined with state-of-the-art (SOTA) Machine Learning (ML) and Deep Learning (DL) algorithms as the core engine is widely used to detect and classify voice phishing call transcripts. While numerous studies on the classification of voice phishing call transcripts are being conducted and demonstrated good performances, with the increase of non-face-to-face financial transactions, there is still the need for improvement using the latest NLP technologies. This paper conducts a benchmarking of Korean voice phishing detection performances of the pre-trained Korean language model KoBERT, against multiple other SOTA algorithms based on the classification of related transcripts from the labeled Korean voice phishing dataset called KorCCVi. The results of the experiments reveal that the classification accuracy on a test set of the KoBERT model outperforms the performances of all other models with an accuracy score of 99.60%.

• Journal of the Korea Society of Computer and Information
• /
• v.22 no.4
• /
• pp.9-16
• /
• 2017

Machine learning has found widespread implementations and applications in many different domains in our life. Logistic regression is a type of classification in machine leaning, and is used widely in many fields, including medicine, economics, marketing and social sciences. In this paper, we present the MapReduce implementation of three existing algorithms, this is, Gradient Descent algorithm, Cost Minimization algorithm and Newton-Raphson algorithm, for logistic regression on RHadoop that integrates R and Hadoop environment applicable to large scale data. We compare the performance of these algorithms for estimation of logistic regression coefficients with real and simulated data sets. We also compare the performance of our RHadoop and RHIPE platforms. The performance experiments showed that our Newton-Raphson algorithm when compared to Gradient Descent and Cost Minimization algorithms appeared to be better to all data tested, also showed that our RHadoop was better than RHIPE in real data, and was opposite in simulated data.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 1996.04a
• /
• pp.30-35
• /
• 1996

Penalty technique perhaps is the most common technique used in the genetic algorithms for constrained optimization problems. In recent years, several techniques have been proposed in the area of evolutionary computation. However, there is no general guideline on designing penalty function and constructing an efficient penalty function is quite problem-dependent. The purpose of the paper is to give a tutorial survey of recent works on penalty techniques used in genetic algorithms and to give a better classification on exisitng works, which may be helpful for revealing the intrinsic relationship among them and for providing some hints for further studies on penalty techniques.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.10
• /
• pp.5079-5099
• /
• 2018

Regarding to the huge number of connections and the large flow of data on the Internet, Intrusion Detection System (IDS) has a difficulty to detect attacks. Moreover, irrelevant and redundant features influence on the quality of IDS precisely on the detection rate and processing cost. Feature Selection (FS) is the important technique, which gives the issue for enhancing the performance of detection. There are different works have been proposed, but a map for understanding and constructing a state of the FS in IDS is still need more investigation. In this paper, we introduce a survey of feature selection algorithms for intrusion detection system. We describe the well-known approaches that have been proposed in FS for IDS. Furthermore, we provide a classification with a comparative study between different contribution according to their techniques and results. We identify a new taxonomy for future trends and existing challenges.

• Journal of information and communication convergence engineering
• /
• v.13 no.2
• /
• pp.113-122
• /
• 2015

Feature subset selection is as a pre-processing step in learning algorithms. In this paper, we propose an efficient algorithm, ModifiedFAST, for feature subset selection. This algorithm is suitable for text datasets, and uses the concept of information gain to remove irrelevant and redundant features. A new optimal value of the threshold for symmetric uncertainty, used to identify relevant features, is found. The thresholds used by previous feature selection algorithms such as FAST, Relief, and CFS were not optimal. It has been proven that the threshold value greatly affects the percentage of selected features and the classification accuracy. A new performance unified metric that combines accuracy and the number of features selected has been proposed and applied in the proposed algorithm. It was experimentally shown that the percentage of selected features obtained by the proposed algorithm was lower than that obtained using existing algorithms in most of the datasets. The effectiveness of our algorithm on the optimal threshold was statistically validated with other algorithms.

• Journal of information and communication convergence engineering
• /
• v.8 no.4
• /
• pp.370-376
• /
• 2010

Ensemble learning is a method for improving the performance of classification and prediction algorithms. However, its performance can be degraded due to multicollinearity problem where multiple classifiers of an ensemble are highly correlated with. This paper proposes genetic algorithm-based optimization techniques of SVM ensemble to solve multicollinearity problem. Empirical results with bankruptcy prediction on Korea firms indicate that the proposed optimization techniques can improve the performance of SVM ensemble.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2002.11a
• /
• pp.779-782
• /
• 2002

본 논문에서는 주성분분석(principal component analysis; 이하 PCA) 및 첨도(Kurtosis)에서의 그레이 블록 거리 알고리즘(grey block algorithms; 이하 GBD)을 이용, 영상간의 거리를 측정하여 어느 정도 영상간의 상대적 식별을 용이하게 하여 영상 분류가 되는지 모의실험을 통하여 확인하고자 한다. 모의실험 결과로부터, PCA에서는 k가 9에서 상대적 식별이 불가능함을 보였고, 첨도에서는 k가 4까지만 블록을 택할 할 수 있음을 모의실험을 통하여 확인할 수 있었다.