Comparison of rough-set and statistical methods in inductive learning

https://doi.org/10.1016/S0020-7373(86)80033-5Get rights and content

Quinlan suggested an inductive algorithm based on the statistical theory of information originally proposed by Shannon. Recently Pawlak showed that the principles of inductive learning (learning from examples) can be precisely formulated on the basis of the theory of rough sets. These two approaches are apparently very different, although in both methods objects in the knowledge base are assumed to be characterized by “features” (attributes and attribute values). The main objective of this paper is to show that the concept of “approximate classification” of a set is closely related to the statistical approach. In fact, in the design of inductive programs, the criterion for selecting dominant attributes based on the concept of rough sets is a special case of the statistical method if equally probable distribution of objects in the “doubtful region” of the approximation space is assumed.

References (15)

  • B.G. Buchanan et al.

    Dendral and metadendral, their applications dimension

    Artificial Intelligence

    (1978)
  • S.M. Weiss et al.

    A model-based method for computer-aided medical decision-making

    Artificial Intelligence

    (1978)
  • R. Davis

    Application of meta level knowledge to the construction, maintenance and use of large knowledge bases

  • T.G. Dietterich et al.

    Learning and generalization of characteristic descriptions: evaluation criteria and comparative review of selected methods

  • R.A. Feigenbaum

    The art of artificial intelligence: I. Themes and case studies of knowledge engineering

  • D.A. McQuarrie
  • R.S. Michalski

    Variable-valued logic: system VL1

There are more references available in the full text version of this article.

Cited by (74)

  • Financial time series forecasting using rough sets with time-weighted rule voting

    2016, Expert Systems with Applications
    Citation Excerpt :

    There are numerous studies comparing the RS approach with other methods. Already Wong, Ziarko, and Ye (1986) have compared the RS approach with statistical methods in machine learning, showing that the RS-based concept of approximate classification is closely related to the statistical approach. Zhong, Dong, and Ohsuga (2001) discusses some disadvantages of two feature selection methods for selecting relevant attributes, namely the filter approach and the wrapper approach, comparing them to the RS approach.

  • Sustainable service and energy provision based on agile rule induction

    2016, International Journal of Production Economics
    Citation Excerpt :

    However, these two approaches are apparently very different. In both methods, objects in the knowledge base are assumed to be characterized by “attributes and attribute values” (Wong et al., 1986). In the decision tree, the rule generated from the complete attribute tree.

  • A comparison of parallel large-scale knowledge acquisition using rough set theory on different MapReduce runtime systems

    2014, International Journal of Approximate Reasoning
    Citation Excerpt :

    Example 1 continued In the earlier studies, Wong et al. used the confidence and resolution factor for inductive learning [42]. Tsumoto proposed the accuracy and coverage to measure the degree of sufficiency and necessity, respectively, and acquired classification rules with high accuracy and high coverage [40].

View all citing articles on Scopus
View full text