A fuzzy search method for rough sets in data mining
Osei Adjei, Li Chen, Heng-Da Cheng, Donald H. Cooley, R.J. Cheng, Xander Twombly · 2002
This paper proposes a technique that combines a fuzzy search method called /spl lambda/-connected search and rough sets, in data mining. /spl lambda/-connected searching was originally proposed to search seismic layers in seismic data processing. Although /spl lambda/-connected searching is designed for digital spaces, or numerical data analysis, it can be used for any domain, as long as the domain can be described by a graph. /spl lambda/-connectedness is an equivalence relation, therefore all searched components form a partition of the base domain. Rough sets, a new methodology in data mining, is based on a classification R on a base set U (the universal set). Then, any subset of U can be represented by an approximation based on the union of certain classes with respect to R. For data processing, U usually is a digital space. The value of each point is often a vector of real/rational numbers. Base domain classification is the key to a rough set system. Theoretically, any equivalence relation R can be defined by a /spl lambda/-connected classification. In order to use the concept of /spl lambda/-connectedness in rough sets and data mining, this paper proposes a limited multi-level /spl lambda/-connected search. In addition, some properties of rough sets using /spl lambda/-connectedness and their applications to data mining are investigated.