Data Mining Based on Evidence Theory
Di Cai · Studies in fuzziness and soft computing · 2001
Data mining is currently one of the most exciting and challenging areas. The concept of linguistic summaries is a user friendly way to express information contained in a database. Commonsense knowledge is a collection of linguistic propositions, that is, propositions with implied imprecise and uncertain quantifiers. The Dempster-Shafer (D-S) theory of evidence fits in handling both imprecision and uncertainty very well. This work uses the D-S theory to establish a framework for dealing with integration of data for distributd databases. Using evidence theory, this work also introduces concept of linguistic summaries and studies their applications to knowledge discovery in distributed databases. We illustrate the use of linguistic summaries by means of running examples using data of risk status conditioned on savings accounts from banks.