Browsing by Author "Bing LIU"

Now showing 1 - 4 of 4
  • No Thumbnail Available
    Item
    AC-5*:An Improved AC-5 and its Specializations
    (1996-04-01T00:00:00Z) Bing LIU
    Many general and specific arc consistency algorithms have been produced in the past for solving Constraint Satisfaction Problems (CSP). The important general algorithms are AC-3, AC-4, AC-5 and AC-6. AC-5 is also a generic algorithm. It can be reduced to AC-3, AC-4 and AC-6. Specific algorithms are efficient specializations of the general ones for specific constraints. Functional, anti-functional and monotonic constraints are three important classes of specific constraints. AC-5 has been specialized to produce an O(ed) algorithm (in time) for these classes of constraints. However, this specialization does not reduce the space requirement. In practical applications, both time and space requirements are important criteria in choosing an algorithm. This paper makes two contributions. First, we propose an improved generic arc consistency algorithm, called AC-5*. It can be specialized to reduce both time and space complexities. Second, we present a more efficient technique for handling an important subclass of functional constraints, namely increasing functional constraints. This technique is significant because in practice almost all functional constraints are actually increasing functional constraints.
  • No Thumbnail Available
    Item
    Finding Interesting Patterns using User Expectations
    (1996-07-01T00:00:00Z) Bing LIU; Wynne HSU; Lai-Fun MUN; Hing-Yan LEE
    Abstract not available.
  • No Thumbnail Available
    Item
    Identifying Interesting Missing Patterns
    (1996-08-01T00:00:00Z) Bing LIU; Wynne HSU; Lai-Fun MUN; Hing-Yan LEE
    Abstract not available.
  • No Thumbnail Available
    Item
    Tuple-Level Analysis for Identification of Interesting Patterns
    (1996-05-01T00:00:00Z) Bing LIU; Wynne HSU; Hing-Yan LEE; Lai-Fun MUN
    One of the important issues in data mining is the "interestingness" problem. Past research and applications have shown that in many situations a huge number of patterns can be discovered from a database. Most of these patterns are actually useless or uninteresting to the user. But because of the huge number of patterns, it is difficult for the user to identify those interesting to him/her. In this project, we propose a new technique to help the user identify interesting patterns. The user is first asked to provide his/her expected patterns according to his/her past knowledge and/or intuitive feelings. Given these expectations, the system uses a tuple-level fuzzy matching technique to analyze and rank the discovered patterns according to a number of interestingness measures. With this technique, the user can quickly focus on a subset of the discovered patterns with the most application values.