• DocumentCode
    1202714
  • Title

    Outlier mining in large high-dimensional data sets

  • Author

    Angiulli, Fabrizio ; Pizzuti, Clara

  • Author_Institution
    Inst. of high Performance Comput. & Networking, Italian Nat. Res. Council, Rende, Italy
  • Volume
    17
  • Issue
    2
  • fYear
    2005
  • Firstpage
    203
  • Lastpage
    215
  • Abstract
    A new definition of distance-based outlier and an algorithm, called HilOut, designed to efficiently detect the top n outliers of a large and high-dimensional data set are proposed. Given an integer k, the weight of a point is defined as the sum of the distances separating it from its k nearest-neighbors. Outlier are those points scoring the largest values of weight. The algorithm HilOut makes use of the notion of space-filling curve to linearize the data set, and it consists of two phases. The first phase provides an approximate solution, within a rough factor, after the execution of at most d + 1 sorts and scans of the data set, with temporal cost quadratic in d and linear in N and in k, where d is the number of dimensions of the data set and N is the number of points in the data set. During this phase, the algorithm isolates points candidate to be outliers and reduces this set at each iteration. If the size of this set becomes n, then the algorithm stops reporting the exact solution. The second phase calculates the exact solution with a final scan examining further the candidate outliers that remained after the first phase. Experimental results show that the algorithm always stops, reporting the exact solution, during the first phase after much less than d + 1 steps. We present both an in-memory and disk-based implementation of the HilOut algorithm and a thorough scaling analysis for real and synthetic data sets showing that the algorithm scales well in both cases.
  • Keywords
    computational complexity; data mining; very large databases; HilOut algorithm; disk-based algorithm; distance-based outlier; in-memory algorithm; large high-dimensional data set; outlier mining; space-filling curve; Accuracy; Algorithm design and analysis; Classification algorithms; Costs; Credit cards; Data mining; Degradation; Medical tests; Medical treatment; Telecommunication services;
  • fLanguage
    English
  • Journal_Title
    Knowledge and Data Engineering, IEEE Transactions on
  • Publisher
    ieee
  • ISSN
    1041-4347
  • Type

    jour

  • DOI
    10.1109/TKDE.2005.31
  • Filename
    1377172