A Fast Granular Method for Classifying Incomplete Inconsistent Data

Abstract

Today extracting knowledge from "inferior quality" data that is characterized by incompleteness and inconsistency is an unavoidable and challenging topic in the field of data mining. In this paper, we propose a fast granular method to classify incomplete inconsistent data using attribute-value block technique. Firstly, a granulation model is constructed to provide a foundation for efficient computation. Secondly, an algorithm of acquiring classification rules is proposed and then an algorithm of minimizing rule sets is proposed, and with these proposed algorithms, a classification algorithm is designed to construct a rule-based classifier. Finally, we use the experiment results to illustrate the effectiveness of the proposed algorithms.