• The KIPS Transactions:PartB
• /
• v.11B no.5
• /
• pp.545-554
• /
• 2004

Combining multiple classifiers to obtain improved performance over the individual classifier has been a widely used technique. The task of constructing a multiple classifier system(MCS) contains two different Issues how to generate a diverse set of base-level classifiers and how to combine their predictions. In this paper, we review the characteristics of existing multiple classifier systems : Bagging, Boosting, and Slaking. For document classification, we propose new MCSs such as Stacked Bagging, Stacked Boosting, Bagged Stacking, Boosted Stacking. These MCSs are a sort of hybrid MCSs that combine advantages of existing MCSs such as Bugging, Boosting, and Stacking. We conducted some experiments of document classification to evaluate the performances of the proposed schemes on MEDLINE, Usenet news, and Web document collections. The result of experiments demonstrate the superiority of our hybrid MCSs over the existing ones.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.37 no.5
• /
• pp.1-10
• /
• 2000

In this paper, we propose a Fuzzy Classifier System(FCS) makes the classifier system be able to carry out the mapping from continuous inputs to outputs. The FCS is based on the fuzzy controller system combined with machine learning. Therefore the antecedent and consequent of a classifier in FCS are the same as those of a fuzzy rule. In this paper, the FCS modifies input message to fuzzified message and stores those in the message list. The FCS constructs rule-base through matching between messages of message list and classifiers of fuzzy classifier list. The FCS verifies the effectiveness of classifiers using Bucket Brigade algorithm. Also the FCS employs the Genetic Algorithms to generate new rules and modify rules when performance of the system needs to be improved. Then the FCS finds the set of the effective rules. We will verify the effectiveness of the poposed FCS by applying it to Autonomous Mobile Robot avoiding the obstacle and reaching the goal.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.14 no.2
• /
• pp.534-539
• /
• 2010

In genomic studies, thousands of features are collected on relatively few samples. One of the goals of these studies is to build classifiers to predict the outcome of future observations. There are three inherent steps to build classifiers: a significant gene selection, model selection and prediction assessment. In the paper, with a focus on prediction assessment, we normalize microarray data with quantile-normalization methods that adjust quartile of all slide equally and then design a system comparing several methods to estimate 'true' prediction error of a prediction model in the presence of feature selection and compare and analyze a prediction error of them. LOOCV generally performs very well with small MSE and bias, the split sample method and 2-fold CV perform with small sample size very pooly. For computationally burdensome analyses, 10-fold CV may be preferable to LOOCV.

• Journal of KIISE:Software and Applications
• /
• v.31 no.7
• /
• pp.949-958
• /
• 2004

race detection using a boosting based algorithm requires a very large size of face and nonface data. In addition, the fact that there always occurs a need for adding additional training data for better detection rates demands an efficient incremental teaming algorithm. In the design of incremental teaming based classifiers, the final classifier should represent the characteristics of the entire training dataset. Conventional methods have a critical problem in combining intermediate classifiers that weight updates depend solely on the performance of individual dataset. In this paper, for the purpose of application to face detection, we present a new method to combine an intermediate classifier with previously acquired ones in an optimal manner. Our algorithm creates a validation set by incrementally adding sampled instances from each dataset to represent the entire training data. The weight of each classifier is determined based on its performance on the validation set. This approach guarantees that the resulting final classifier is teamed by the entire training dataset. Experimental results show that the classifier trained by the proposed algorithm performs better than by AdaBoost which operates in batch mode, as well as by ${Learn}^{++}$.

• Journal of Information Processing Systems
• /
• v.16 no.1
• /
• pp.6-29
• /
• 2020

Biometrics identification using multiple modalities has attracted the attention of many researchers as it produces more robust and trustworthy results than single modality biometrics. In this paper, we present a novel multimodal recognition system that trains a deep learning network to automatically learn features after extracting multiple biometric modalities from a single data source, i.e., facial video clips. Utilizing different modalities, i.e., left ear, left profile face, frontal face, right profile face, and right ear, present in the facial video clips, we train supervised denoising auto-encoders to automatically extract robust and non-redundant features. The automatically learned features are then used to train modality specific sparse classifiers to perform the multimodal recognition. Moreover, the proposed technique has proven robust when some of the above modalities were missing during the testing. The proposed system has three main components that are responsible for detection, which consists of modality specific detectors to automatically detect images of different modalities present in facial video clips; feature selection, which uses supervised denoising sparse auto-encoders network to capture discriminative representations that are robust to the illumination and pose variations; and classification, which consists of a set of modality specific sparse representation classifiers for unimodal recognition, followed by score level fusion of the recognition results of the available modalities. Experiments conducted on the constrained facial video dataset (WVU) and the unconstrained facial video dataset (HONDA/UCSD), resulted in a 99.17% and 97.14% Rank-1 recognition rates, respectively. The multimodal recognition accuracy demonstrates the superiority and robustness of the proposed approach irrespective of the illumination, non-planar movement, and pose variations present in the video clips even in the situation of missing modalities.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.1
• /
• pp.110-114
• /
• 2010

We describe a new method for text categorization using TextRank algorithm. Text categorization is a problem that over one pre-defined categories are assigned to a text document. TextRank algorithm is a graph-based ranking algorithm. If we consider that each word is a vertex, and co-occurrence of two adjacent words is a edge, we can get a graph from a document. After that, we find important words using TextRank algorithm from the graph and make feature which are pairs of words which are each important word and a word adjacent to the important word. We use classifiers: SVM, Na$\ddot{i}$ve Bayesian classifier, Maximum Entropy Model, and k-NN classifier. We use non-cross-posted version of 20 Newsgroups data set. In consequence, we had an improved performance in whole classifiers, and the result tells that is a possibility of TextRank algorithm in text categorization.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.12 no.12
• /
• pp.2357-2364
• /
• 2008

In this paper, we consider the problem of exploiting a taxonomy of propositionalized attributes in order to generate compact and robust classifiers. We introduce Propositionalized Attribute Taxonomy guided Naive Bayes Learner (PAT-NBL), an inductive learning algorithm that exploits a taxonomy of propositionalized attributes as prior knowledge to generate compact and accurate classifiers. PAT-NBL uses top-down and bottom-up search to find a locally optimal cut that corresponds to the instance space from propositionalized attribute taxonomy and data. Our experimental results on University of California-Irvine (UCI) repository data set, show that the proposed algorithm can generate a classifier that is sometimes comparably compact and accurate to those produced by standard Naive Bayes learners.

• Korean Journal of Remote Sensing
• /
• v.19 no.2
• /
• pp.170-179
• /
• 2003

In order to classify Land Use/Land Cover using multispectral images, we have to give consequence to defining proper classes and selecting training sample with higher class separability. The process of satellite hyperspectral image which has a lot of bands is difficult and time-consuming. Furthermore, classification result of hyperspectral image with noise is often worse than that of a multispectral image. When selecting training fields according to the signatures in the study area, it is difficult to calculate covariance matrix in some clusters with pixels less than the number of bands. Therefore in this paper we presented an overview of feature extraction methods for classification of Hyperion data and examined effectiveness of feature extraction through the accuracy assesment of classified image. Also we evaluated the classification accuracy of optimal meaningful features by class separation distance, which is also a method for band reduction. As a result, the classification accuracies of feature-extracted image and original image are similar regardless of classifiers. But the number of bands used and computing time were reduced. The classifiers such as MLC, SAM and ECHO were used.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2017.05a
• /
• pp.370-370
• /
• 2017

최근 기후변화와 이상기후의 영향으로 인하여 홍수재해의 시 공간적 패턴은 보다 복잡해지고, 예측이 어려워지고 있다. 이러한 기상이변에 따른 홍수피해를 예방하기 위한 비구조적 대책으로 홍수위험등급 및 범람범위 등의 정보를 포함하고 있는 홍수위험지도의 작성이 필요하다. 실제로 고정밀도 홍수위험지도를 작성하기 위해서는 지형, 지질, 기상 등의 디지털 정보 및 사회 경제와 관련된 다양한 DB를 필요로 하며, 강우-유출-범람해석 모델링을 통해 범람면적 및 침수깊이 등의 정보를 획득하게 된다. 하지만 일부지역, 특히 개발도상국에서는 이러한 계측 홍수 데이터가 부족하거나 획득할 수가 없어 홍수위험지도 제작이 불가능하거나 그 정확도가 매우 낮은 실정이다. 따라서 본 연구에서는 ASTER 또는 SRTM과 같은 범용 DEM 등 지형자료만을 기반으로 한 선형이진분류법(Liner binary classifiers)과 ROC분석(Receiver Operation Characteristics)을 이용하여 미계측 유역 (DB부재 또는 부족으로 강우-유출-범람해석 모델링이 불가능한 북한지역)의 홍수위험지역을 탐지하고, 적용성을 평가하고자 한다. 5개의 단일 지형학적 지수와 6개의 복합 지형학적 지수를 이용하여 Area Under the Curve (AUC)를 계산하고, Sensitivity (민감도)와 Specificity (특이도)가 가장 높은 지수를 선별하여 홍수위험지도를 작성하고, 실제 홍수범람 영상(2007년 북한 함경남도지역 용흥강 홍수)과 비교 분석하였다. 본 연구에서 제시하는 선형이진분류법과 ROC분석 방법은 홍수범람해석을 위한 다양한 기초정보를 필요로 하지 않고, 지형정보만을 사용하기 때문에 관측 데이터가 없거나 부족한 지역에 대해서 우선적으로 홍수위험지역을 탐지하고, 선별하는데 유용할 것으로 판단된다.

• Journal of the Korea Society of Computer and Information
• /
• v.18 no.8
• /
• pp.17-25
• /
• 2013

TIn this paper, we propose a model automatically scoring a student's answer for a descriptive problem by using a voting method. Considering the model construction cost, the proposed model does not separately construct the automatic scoring model per problem type. In order to utilize features useful for automatically scoring the descriptive answers, the proposed model extracts feature values from the results, generated by comparing the student's answer with the answer sheet. For the purpose of improving the precision of the scoring result, the proposed model collects the scoring results classified by a few machine learning based classifiers, and unanimously selects the scoring result as the final result. Experimental results show that the single machine learning based classifier C4.5 takes 83.00% on precision while the proposed model improve the precision up to 90.57% by using three machine learning based classifiers C4.5, ME, and SVM.