Entity

Time filter

Source Type

Yanggu, South Korea

Kang H.-S.,Korea University | Nam K.,UMLogics Co. | Kim S.-I.,Korea University
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2012

As textual data have exponentially increased, it is focused that a need for automatic classification of relevant data to one of pre-defined classes. In many practical applications, they assume that training data are evenly distributed among all classes, but they are suffered from an imbalanced problem. Several algorithms and re-sampling methods have been proposed to overcome an imbalanced problem, but they are still facing the overfitting and information missing. This paper proposes the Decomposed K-Nearest Neighbor (DCM-KNN). In training step, the DCM-KNN decomposes training data into misclassified and correctly-classified data set based on the result of traditional KNN, and finds the appropriate KNN for each set. In test step, the DCM-KNN estimates whether test data is similar to misclassified and correctly-classified data set, and applies the appropriate KNNs. Experimental results show that proposed algorithm can achieve more accurate results in an imbalanced condition. © 2012 Springer-Verlag. Source


An S.-H.,Daejeon University | Nam K.,UMLogics Co. | Jeong M.-K.,UMLogics Co. | Choi Y.-R.,Daejeon University
International Journal of Multimedia and Ubiquitous Engineering | Year: 2014

In order to preserve important data including a company’s intellectual properties which require security or deserve to be archives, we generally do it electronically in disk mirroring, also known as RAID, or external storage devices. Digitized data can be corrupted if the storage devices deteriorate or by external access, and such compromised data undermine their authenticity and usability. This thesis presents a new method to recover damaged electronic data to restore their authenticity and usability. Copyright © 2014 SERSC Source


An S.-H.,Daejeon University | Nam K.,UMLogics Co. | Jeong M.-K.,UMLogics Co. | Choi Y.-R.,Daejeon University
International Journal of Security and its Applications | Year: 2016

This thesis proposes a method to detect sophisticated electronic financial frauds using SVDD. The financial industry detects electronic financial frauds using FDS, but its false positive rate is high enough to require additional authentications. It causes customers inconveniences and does not detect those sophisticated financial frauds. In order to resolve the aforementioned issues, this study proposes a method to detect such potential frauds by profiling and vectorizing user activities and device information by SVDD. © 2016 SERSC. Source


Jeong M.-K.,UMLogics Co. | An S.-H.,UMLogics Co. | Nam K.,UMLogics Co.
International Journal of Security and its Applications | Year: 2016

This thesis proposes a method to detect financial fraudby dividing users' financial transactions into a normal area and an abnormal area,using SVDD and train the areas as such fraud evolves in terms of complexity. The existing financial industry detects electronic financial frauds using FDS, but its false positive rate is high enough to require additional authentications of user information. It causes customers inconveniences and does not always detect those sophisticated financial frauds. In order to resolve the aforementioned issues, this study proposes a method to detect such potential frauds by profiling user financial transaction data including user activities, device information, andtransaction patterns and vectorizing them into a normal area and an abnormal area using SVDD. © 2016 SERSC. Source


Jeong Y.,UMLogics Co. | Nam K.,UMLogics Co. | Jeong M.K.,UMLogics Co.
International Journal of Software Engineering and its Applications | Year: 2014

This thesis presents an improved algorithm of ν-SVDD that detects outliers in documents. The existing algorithms produce a virtual modulation vector with all elements of a power set of a feature vector from a document, and the virtual modulation vector may include vectors that do not reflect true characteristics of the document. To address the issue, this study proposes an enhanced version of the ν-SVDD algorithm that produces a more precise virtual modulation vector. Test results run by this new algorithm in this paper show excellent precision in detecting leakage of tampered documents. © 2014 SERSC. Source

Discover hidden collaborations