• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1996년도 하계학술대회 논문집 B
• /
• pp.859-861
• /
• 1996

This paper presents the novel fault diagnostic expert system based on dissolved gas analysis(DGA) techniques in power transformer. The uncertainty of key gas analysis, norm threshold, and gas ratio boundaries are managed by using a fuzzy set concept. The uncertainty of rules are handled by fuzzy measures. Trend analysis through the monthly increment of key gas and DGA analysis are combined by the Dempster-Shafer theory, and the state of transformer and confidence factor are yielded by using this combined analysis. To verify the effectiveness of the proposed diagnosis technique, the expert system has been tested by using KEPCO's transformer gas records.

• International journal of advanced smart convergence
• /
• 제8권2호
• /
• pp.94-101
• /
• 2019

Building a stable knowledge base is an important issue in the application of knowledge engineering. In this paper, we present an algorithm for detecting and locating discrepancies in the line of the reasoning process especially when discrepancies occur on belief values. This includes backtracking the rule firing from a goal node of the rule network. Retracting a belief function allows the current belief state to move back to another belief state without the rule firing. It also gives an estimate, called contribution measure, of how much the rule has an impact on the current belief state. Examining the measure leads the expert to locate the possible cause of problem in the rule. For non-monotonic reasoning, the belief retraction method moves the belief state back to the previous state. A tracing algorithm is presented to identify and locate the cause of problem. This also gives repair suggestions for rule refinement.

• International journal of advanced smart convergence
• /
• 제8권1호
• /
• pp.133-140
• /
• 2019

In this paper, we present an algorithm for adjusting degree of belief for consistency on the evidence dependency network where various sets of evidence support different sets of hypotheses. It is common for experts to assign higher degree of belief to a hypothesis when there is more evidence over the hypothesis. Human expert without knowledge of uncertainty handling may not be able to cope with how evidence is combined to produce the anticipated belief value. Belief in a hypothesis changes as a series of evidence is known to be true. In non-monotonic reasoning environments, the belief retraction method is needed to clearly deal with uncertain situations. We create evidence dependency network from rules and apply the evidence retraction algorithm to refine belief values on the hypothesis set. We also introduce negative belief values to reflect the reverse effect of evidence combination.

• 한국국방경영분석학회지
• /
• 제23권2호
• /
• pp.45-71
• /
• 1997

There has been considerable research recently on uncertainty handling in the fields of artificial intelligence and knowledge-based system. Various numerical and non-numerical methods have been proposed for representing and propagating uncertainty in knowledge-based system. The Bayesian method, the Dempster-Shafer's Evidence Theory, the Certainty Factor model and the Fuzzy Set Theory are most frequently appeared in the knowledge-based system. Each of these four methods views uncertainty from a different perspective and propagates it differently. There is no single method which can handle uncertainty properly in all kinds of knowledge-based systems' domain. Therefore a knowledge-based system will work more effectively when the uncertainty handling method in the system fits to the system's environment. This paper proposed a framework for selecting proper uncertainty handling methods in knowledge-based system with respect to characteristics of problem domain and cognitive styles of experts. A schema with strategic/operational and unstructured/structured classification is employed to differenciate domain. And a schema with systematic/intuitive and preceptive/receptive classification is employed to differenciate experts' cognitive style. The characteristics of uncertainty handling methods are compared with characteristics of problem domains and cognitive styles respectively. Then a proper uncertainty handling method is proposed for each category.

• 정보처리학회논문지B
• /
• 제9B권4호
• /
• pp.389-390
• /
• 2002

동적 웹 컨텐츠 제공에서 고객을 위한 추천서비스에 이르는 인터넷 기반의 전자상거래 애플리케이션에서는 고객이 어떤 성향을 가지고 있는가에 대한 정보를 획득하는 것이 중요하다. 웹 개인화의 대표적인 기술인 협력적 석과는 사용자의 정보를 정적인 프로파일 형태로 저장하여 사용자의 성향 변화를 빨리 획득할 수 없다. 또한 사용자의 명시적 평가 의존성, 확장성 부족, 다차원 공간 데이터에 대한 적용 어려움 둥의 문제점을 가지고 있다. 이와 같은 단점을 해결하기 위한 해결 방안으로 웹 사용 정보 마이닝(web usage mining)이 쓰이고 있다. 웹 사용 정보 마이닝은 서버에 축적된 웹 사용 데이터(web usage data)를 이용하여 패턴을 발견하는 기술이다. 특히 연관 규칙 생성 알고리즘으로 웹 사용 패턴(web usage pattern)을 찾고 패턴을 클러스터링하는 기술이 사용되고 있다. 그러나 연관 규칙 생성 알고리즘은 많은 수의 패턴들을 찾고 또 유용하지 못한 패턴을 발견하는 단점이 있다. 본 논문에서는 검증된 웹 사용 패턴을 이용한 동적 사용자 프로파일 생성 방법을 제안한다. 먼저 패턴 발견을 위해 연관 규칙 생성 알고리즘인 Apriori를 이용하고 사용자 프로파일을 위한 클러스터를 생성하기 위해 ARHP를 채택하였다. 클러스터를 생성하기 전에 Dempster-Shafer 이론을 이용하여 유용하지 못한 패턴을 제거하는 패턴 검증 과정을 수행한다. 검증된 패턴을 이용하여 클러스터를 생성하고 사용자의 현재 활성화된 세션에 따라 동적으로 사용자 프로파일이 생성된다

• 대한원격탐사학회지
• /
• 제28권6호
• /
• pp.611-622
• /
• 2012

This study proposed two multisensor fusion methods for segment-based image classification utilizing a region-growing segmentation. The proposed algorithms employ a Gaussian-PDF measure and an evidential measure respectively. In remote sensing application, segment-based approaches are used to extract more explicit information on spatial structure compared to pixel-based methods. Data from a single sensor may be insufficient to provide accurate description of a ground scene in image classification. Due to the redundant and complementary nature of multisensor data, a combination of information from multiple sensors can make reduce classification error rate. The Gaussian-PDF method defines a regional measure as the PDF average of pixels belonging to the region, and assigns a region into a class associated with the maximum of regional measure. The evidential fusion method uses two measures of plausibility and belief, which are derived from a mass function of the Beta distribution for the basic probability assignment of every hypothesis about region classes. The proposed methods were applied to the SPOT XS and ENVISAT data, which were acquired over Iksan area of of Korean peninsula. The experiment results showed that the segment-based method of evidential measure is greatly effective on improving the classification via multisensor fusion.

• 정보관리학회지
• /
• 제8권1호
• /
• pp.100-116
• /
• 1991

본 연구에서는 지식베이스의 용어간 관계와 증거추론이론을 이용한 정보검색 시스템을 설계하였다. 실험을 위한 문헌파일은 한국전자통신연구소의 통신분야 기술문서를 대상으로 작성하였으며, 지식베이스는 INSPEC 시소러스의 통신분야 용어 및 용어간 관계들을 발췌하여 구성하였다. 그리고 용어간의 관련성은 용어간 관계의 종류에 따라 수치로 표현하였으며, 이들 수치를 이용하여 뎀스터-셰이퍼 이론에 따라 질문과 문헌간의 관련성을 추론, 산출하였다. 실험결과 질문의 탐색어외에도 관련된 용어의 확장검색을 통하여 포괄적인 검색을 할 수 있었으며, 용어간 관계를 반영하여 질문과 문헌간의 관련성을 산출하고, 관련성 순위에 따라 검색결과를 제시할 수 있었다.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2001년도 하계학술대회 논문집 D
• /
• pp.2744-2746
• /
• 2001

이 논문에서는 이동 로봇을 위하여 퍼지이론과 Dempster-Shafer 이론을 이용한 불확실한 환경에서의 센서기반 네비게이션 방법을 제안한다. 제안된 제어기는 장애물 회피 동작과 목적지 찾기 동작을 위한 2개의 행동 모듈로 구성되어 있다. 2개의 행동 모듈은 각각 퍼지 이론으로 학습되었고, 적절한 행동 선택 방법으로 선택되게끔 하였다. 견고한 퍼지 제어기를 가진 로봇이 실험 환경내에서 안전하게 움직이기 위하여 자동으로 지도를 구축(Map Building) 하도록 하였다. 이 실험에서 구성된 맵은 평면상의 격자를 중심으로 작성되었고 로봇의 센서에서 읽어들인 센서 값은 D-S 추론 이론을 이용하여 기존의 맵과 혼합되어진다. 즉, 로봇이 움직일때 마다 실험 환경내에서 새로운 정보를 읽어 들이고, 그 정보로 인하여 기존의 지도가 새로운 지도로 갱신되는 것이다. 이러한 작업을 거치면서 로봇은 장애물과 충돌없이 배회하는 것 뿐 아니라 설정된 목적지까지도 쉽게 찾아갈 수가 있다. 실험에 대한 안정성과 확신을 검증 받기 위하여 실제 로봇에 적용하기보다는 먼저 이동 로봇의 시뮬레이션으로 실험 해 보고자 한다.

• 한국항만학회지
• /
• 제12권1호
• /
• pp.9-17
• /
• 1998

The evaluation structure of complex problems is composed of multi-attributes and hierarchy. A many studies were existed on this problems, but that based on the assumption that the evaluation elements were independent. The actual evaluation problems have the complexity, ambiguity and interlinkage among the elements. In this situation, the fuzzy evaluation process is very effective in settling the complex problems. For evaluation of large scale hierarchical MADM problem, the fuzzy evaluation algorithm is developed in this paper, and that is centering on the identification of fuzzy measures. In this study, we newly identified the weight and interaction among the evaluation attributes. The results of this study are as follows: we can identified the hierarchical structure of the evaluation problem which is composed of the evaluation structure, function and hierarchy; we improved the existed weighting method which could be accomplished by normalizing process, considering the uncertainty and new weight integrating method which come from Dempster-Shafer theory. And we take into account the interaction properties among more than 3 evaluation attributes, which can be compared with the existed studies in which only 2 evaluation attributes taked into account.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권10호
• /
• pp.3965-3982
• /
• 2015

Event detection is one of the key issues in many wireless sensor network (WSN) applications. The uncertainties that are derived from the instability of sensor node, measurement noise and incomplete sampling would influence the performance of event detection to a large degree. Many of the present researches described the sensor readings with crisp values, which cannot adequately handle the uncertainties inhered in the imprecise sensor readings. In this paper, a fault-tolerant event detection algorithm is proposed based on Dempster-Shafer (D-S) theory (also called evidence theory). Instead of crisp values, all possible states of the event are represented by the Basic Probability Assignment (BPA) functions, with which the output of each sensor node are characterized as weighted evidences. The combination rule was subsequently applied on each sensor node to fuse the evidences gathered from the neighboring nodes to make the final decision on whether the event occurs. Simulation results show that even 20% nodes are faulty, the accuracy of the proposed algorithm is around 80% for event region detection. Moreover, 97% of the error readings have been corrected, and an improved detection capability at the boundary of the event region is gained by 75%. The proposed algorithm can enhance the detection accuracy of the event region even in high error-rate environment, which reflects good reliability and robustness. The proposed algorithm is also applicable to boundary detection as it performs well at the boundary of the event.