본문 바로가기
HOME> 보고서 > 보고서 검색상세

보고서 상세정보

Development of Algorithms for Anomaly Detection and Dynamic Centrality Measures with the Applications to Identifying Emerging Technologies
Development of Algorithms for Anomaly Detection and Dynamic Centrality Measures with the Applications to Identifying Emerging Technologies

  • 주관연구기관

    Rutgers, The State University of New Jersey
    Rutgers, The State University of New Jersey

  • 연구책임자

    정명기

  • 참여연구자

    Ali Tosyali   최정섭   김진호  

  • 보고서유형

    최종보고서

  • 발행국가

    대한민국

  • 언어

    대한민국

  • 발행년월

    2015-10

  • 주관부처

    미래창조과학부
    KA

  • 사업 관리 기관

    한국과학기술정보연구원
    Korea Institute of Science and Technology Information

  • 등록번호

    TRKO201600000538

  • DB 구축일자

    2016-04-16

  • 초록 


    Chapter 1
    Introduction
    In recent years, there has been an emphasis on analyzing data using graph theoretical methods (Easle...

    Chapter 1
    Introduction
    In recent years, there has been an emphasis on analyzing data using graph theoretical methods (Easley & Kleinberg, 2010 and van Steen, 2010). Graph-based data mining approaches attempt to analyze data that can be represented in a graph, consisting of nodes and edges. Graph ata appears widely in areas such as social networks, brain connectivity graphs, transportation networks, the Internet and other fields. While there has been much work on graph-based data mining (Cook & Holder, 2006, Xu et al., 2007, Holder & Cook, 2009 and Kang et. al, 2013), there is still much room for contribution in the area of graph-based outlier ranking and detection.
    Outlier detection (or anomaly detection) has to do with identifying entities that are unusual or that deviate from the rest of the dataset (Barnett & Lewis, 1994 and Hodge & Austin, 2004). This is an important research topic that has been researched within diverse areas and application domains (Barnett & Lewis, 1994, Hodge & Austin, 2004, Arias-Castro et al., 2011 and Ranshous et al, 2014). Many anomaly detection techniques have been specifically developed for certain application domains, while others are more generic (Chandolar et al., 2009). Researchers have adopted concepts from diverse disciplines such as statistics, machine learning, data mining, information theory, spectral theory, and have applied them to specific problem formulations. The goal of graph outlier ranking is to score and rank objects to the degree that they differ from majority of dataset in the graph data. That is, node relationship data is analyzed to identify interesting or exceptional objects. From an abstract level, an anomaly or outlier is defined as an object that does not conform to expected normal behavior (Chandolar et al., 2009). A straightforward anomaly detection approach, then, is to define a region or characteristic representing normal behavior and identify any observation in the data that does not belong to this normal region as an anomaly.


    ...


  • 목차(Contents) 

    1. COVER ... 1
    2. Table of Contents ... 5
    3. Chapter 1 Introduction ... 7
    4. Chapter 2 Patent Clustering and Outlier Ranking Methodologies for Attributed Patent Citation Networks ... 1...
    1. COVER ... 1
    2. Table of Contents ... 5
    3. Chapter 1 Introduction ... 7
    4. Chapter 2 Patent Clustering and Outlier Ranking Methodologies for Attributed Patent Citation Networks ... 11
    5. 2.1 Introduction ... 11
    6. 2.2 Background ... 12
    7. 2.2.1 Existing methods for node outlier ranking ... 13
    8. 2.2.2 Existing methods for graph structure and node attribute-based node outlierranking ... 14
    9. 2.2.3 Subspace clustering for outlier detection ... 16
    10. 2.3 New subspace clustering algorithm for patents in a patent citation network ... 18
    11. 2.3.1 Characteristics of patent citation networks ... 18
    12. 2.3.2 New subspace clustering algorithm for PCN ... 19
    13. 2.3.3 Subspace clustering numerical example ... 23
    14. 2.4 New node outlier ranking methods for PCNs ... 24
    15. 2.4.1 Integrated graph structure-based and node attribute model ... 25
    16. 2.4.2 Weighted subspace clustering ... 26
    17. 2.4.3 Graph structure-based methods ... 29
    18. 2.5 Experimental results ... 30
    19. 2.5.1 Data description ... 30
    20. 2.5.2 Artificial dataset 1: 6-node attributed patent citation network ... 31
    21. 2.5.3 Artificial dataset 2: 14-node attributed patent citation network ... 33
    22. 2.5.4 Artificial dataset 3: 14-node patent citation network ... 36
    23. 2.5.5 Real-life patent citation network ... 38
    24. Chapter 3 Patent Clustering and Outlier Ranking on PCNs using Nonnegative Matrix Factorization ... 40
    25. 3.1 Introduction ... 40
    26. 3.2 Nonnegative matrix factorization ... 41
    27. 3.3 New scoring method for anomaly detection in directed graph data ... 43
    28. 3.3.1 Weighted adjacency matrix of a patent citation network ... 43
    29. 3.3.2 ANMF with weighted adjacency matrix ... 44
    30. 3.3.3 Scoring method for anomaly detection ... 45
    31. 3.3.4 Initialization based on modified SVD ... 48
    32. 3.4 An Illustrative Example ... 49
    33. Chapter 4 Patent Importance Scoring Methodologies for Dynamic Patent Citation Network ... 53
    34. 4.1 Introduction ... 53
    35. 4.2 Existing methods for measuring importance score in dynamic graphs ... 54
    36. 4.2.1 CiteRank ... 54
    37. 4.2.2 Centrality metric for dynamic networks ... 59
    38. 4.2.3 Time-aware ranking in dynamic citation networks ... 66
    39. 4.3 New scoring method for importance of patents in dynamic graph data ... 68
    40. 4.3.1 Proposed importance measure ... 69
    41. 4.3.2 Experimental evaluation ... 70
    42. Chapter 5 Conclusion and Future Work ... 72
    43. 5.1 Conclusion ... 72
    44. 5.2 Future work ... 73
    45. Reference ... 75
    46. End of Page ... 79
  • 참고문헌

    1. 전체(0)
    2. 논문(0)
    3. 특허(0)
    4. 보고서(0)

 활용도 분석

  • 상세보기

    amChart 영역
  • 원문보기

    amChart 영역