Knowledge Reduction in Random Incomplete Decision Tables via Evidence Theory

Abstract

Statisticians and database users often encounter the problem of missing or imprecise data obtained by a random experiment. Such a data set is called a random incomplete information table. In this paper, we study knowledge reduction in random incomplete information tables and random incomplete decision tables by using a hybrid model based on the rough set theory and the Dempster-Shafer theory of evidence. The concepts of random belief reducts and random plausibility reducts in random incomplete information tables and random incomplete decision tables are introduced. The relationships among the lower approximation reduct, the upper approximation reduct, the random belief reduct, the random plausibility reduct, and the classical reduct in random incomplete decision tables are examined.