• The KIPS Transactions:PartD
• /
• v.19D no.4
• /
• pp.263-270
• /
• 2012

With the current development of high speed Internet and massive database technology, the amount of web documents increases rapidly, and thus, classifying those documents automatically is getting important. In this study, we propose an effective method to extract document features based on Hangeul morpheme and keyword analyses, and to classify non-structured documents automatically by predicting subjects of those documents. To extract document features, first, we select terms using a morpheme analyzer, form the keyword set based on term frequency and subject-discriminating power, and perform the scoring for each keyword using the discriminating power. Then, we generate the classification model by utilizing the commercial software that implements the decision tree, neural network, and SVM(support vector machine). Experimental results show that the proposed feature extraction method has achieved considerable performance, i.e., average precision 0.90 and recall 0.84 in case of the decision tree, in classifying the web documents by subjects.

• KIPS Transactions on Software and Data Engineering
• /
• v.3 no.2
• /
• pp.87-92
• /
• 2014

In this paper, we present a two-phase activity recognition system using smartphone's accelerometers. To consider the unique temporal pattern of accelerometer data for each activity, our system executes the decision-tree(DT) learning in the first phase, and then, in the second phase, executes the hidden Markov model(HMM) learning based on the sequences of classification results of the first phase classifier. Moreover, to build a robust recognizer for each activity, we trained our system using a large amount of data collected from different users, different positions and orientations of smartphone. Through experiments using 6720 examples collected for 6 different indoor activities, our system showed high performance based on its novel design.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.38A no.1
• /
• pp.1-9
• /
• 2013

This paper propose a human action recognition method that uses bag-of-features (BoF) based on CS-LBP (center-symmetric local binary pattern) and a spatial pyramid in addition to the random forest classifier. To construct the BoF, an image divided into dense regular grids and extract from each patch. A code word which is a visual vocabulary, is formed by k-means clustering of a random subset of patches. For enhanced action discrimination, local BoF histogram from three subdivided levels of a spatial pyramid is estimated, and a weighted BoF histogram is generated by concatenating the local histograms. For action classification, a random forest, which is an ensemble of decision trees, is built to model the distribution of each action class. The random forest combined with the weighted BoF histogram is successfully applied to Standford Action 40 including various human action images, and its classification performance is better than that of other methods. Furthermore, the proposed method allows action recognition to be performed in near real-time.

• Journal of the Korea Convergence Society
• /
• v.12 no.2
• /
• pp.7-13
• /
• 2021

This paper introduces an algorithm that compensates for missing values after converting them into a format that can represent the probability for incomplete data including missing values in training data. In the previous method using this data conversion, incomplete data was processed by allocating missing values with an equal probability that missing variables can have. This method applied to many problems and obtained good results, but it was pointed out that there is a loss of information in that all information remaining in the missing variable is ignored and a new value is assigned. On the other hand, in the new proposed method, only complete information not including missing values is input into the well-known classification algorithm (C4.5), and the decision tree is constructed during learning. Then, the probability of the missing value is obtained from this decision tree and assigned as an estimated value of the missing variable. That is, some lost information is recovered using a lot of information that has not been lost from incomplete learning data.

• Annual Conference on Human and Language Technology
• /
• 2020.10a
• /
• pp.563-568
• /
• 2020

본 연구에서는 블랙박스로 알려진 딥러닝 모델에 설명 근거를 제공할 수 있는 설명자 모델을 적용해 보았다. 영화평 감성 분석을 위해 MLP, CNN으로 구성된 딥러닝 모델과 결정트리의 앙상블인 Gradient Boosting 모델을 이용하여 감성 분류기를 구축하였다. 설명자 모델로는 기울기(gradient)을 기반으로 하는 IG와 레이어 사이의 가중치(weight)을 기반으로 하는 CAM, 그리고 설명가능한 대리 모델을 이용하는 LIME과 입력 속성에 대한 선형모델을 추정하는 SHAP을 사용하였다. 설명자 모델의 특성을 보기 위하여 히트맵과 관련성 높은 N개의 속성을 추출해 보았다. 설명자가 제공하는 기여도에 따라 입력 속성을 제거해 가며 분류기 성능 변화를 측정하는 정량적 평가도 수행하였다. 또한, 사람의 판단 근거와의 일치도를 살펴볼 수 있는 '설명 근거 정확도'라는 새로운 평가 방법을 제안하여 적용해 보았다.

• Journal of KIISE
• /
• v.44 no.5
• /
• pp.553-558
• /
• 2017

With the rapid growth of Big Data, research on extracting meaningful information is being pursued by both academia and industry. Especially, data characteristics derived from analysis, and researcher intention are key factors for search algorithms to obtain accurate output. Therefore, reflecting both data characteristics and researcher intention properly is the final goal of data analysis research. The data analyzed properly can help users to increase loyalty to the service provided by company, and to utilize information more effectively and efficiently. In this paper, we explore various methods of document-evaluation, so that we can improve the accuracy of searching article one of the most frequently searches used in real life. We also analyze the experiment result, and suggest the proper manners to use various methods.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2005.11a
• /
• pp.47-50
• /
• 2005

센서와 모바일 기술의 발달로 인해 다양한 센서에서 수집된 스트림 데이터를 처리하는 연구들이 많이 수행되고 있다. 다차원 속성의 스트림 데이터는 센서에서 주기적으로 수집되어 버퍼링 후 처리되기 때문에 기존의 투플 기반의 데이터 분류 기법에 적합하지 않다. 따라서 이 논문에서는 윈도우 기반의 스트림 데이터 분류를 위해 각 속성의 평균과 표준편차 값을 이용하여 투플 기반으로 변환하는 기법을 제안한다. 제안된 기법의 타당성은 투플 기반 데이터 분류 기법(의사결정트리, 단순 베이지안 분류기, 베이지안 신뢰 네트워크)에 의한 정확도 측정에 기반 한다. 로봇에서 수집된 센서 데이터를 이용한 실험 결과, 높은 정확도로 제안된 기법이 타당함을 증명하였으며 베이지안 신뢰 네트워크 기법이 다른 기법에 비해 우수함을 발견하였다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.14 no.9
• /
• pp.1979-1986
• /
• 2010

Biological early warning system detects toxicity by looking at behavior of organisms in water. The system uses classifier for judgement about existence and amount of toxicity in water. Boosting algorithm is one of possible application method for improving performance in a classifier. Boosting repetitively change training example set by focusing on difficult examples in basic classifier. As a result, prediction performance is improved for the events which are difficult to classify, but the information contained in the events which can be easily classified are discarded. In this paper, an incremental learning method to overcome this shortcoming is proposed by using the extended data expression. In this algorithm, decision tree classifier define class distribution information using the weight parameter in the extended data expression by exploiting the necessary information not only from the well classified, but also from the weakly classified events. Experimental results show that the new algorithm outperforms the former Learn++ method without using the weight parameter.

• Journal of KIISE:Software and Applications
• /
• v.29 no.3
• /
• pp.192-201
• /
• 2002

Automated text classification is considered as an important method to manage and process a huge amount of documents in digital forms that are widespread and continuously increasing. Recently, text classification has been addressed with machine learning technologies such as k-nearest neighbor, decision tree, support vector machine and neural networks. However, only few investigations in text classification are studied on real problems but on well-organized text corpus, and do not show their usefulness. This paper proposes and analyzes text classification methods for a real application, email document classification task. First, we propose a combining method of multiple neural networks that improves the performance through the combinations with maximum and neural networks. Second, we present another strategy of combining multiple machine learning classifiers. Voting, Borda count and neural networks improve the overall classification performance. Experimental results show the usefulness of the proposed methods for a real application domain, yielding more than 90% precision rates.

• The KIPS Transactions:PartB
• /
• v.9B no.6
• /
• pp.759-764
• /
• 2002

Although image occupies a large part of importance on the Web documents, there have not been many researches for analyzing and understanding it. Many Web images are used for carrying important information but others are not used for it. In this paper classify the Web images from presently served Web sites to erasable or non-erasable classes. based on machine learning methods. For this research, we have detected 16 special and rich features for Web images and experimented by using the Baysian and decision tree methods. As the results, F-measures of 87.09%, 82.72% were achived for each method and particularly, from the experiments to compare the effects of feature groups, it has proved that the added features on this study are very useful for Web image classification.