A New Approach on Rule Mining Based on Granularity in Incomplete Information Systems

Abstract

The analysis of incomplete decision table is an important research topic in the field of intelligent information processing. This paper defines incomplete information system, incomplete decision table, granularity, tolerance granulation, deterministic operator, possible operator. Firstly, it extracts tolerance granulations of the object collections that are divided by decision attributes. Secondly, it gets the new object collections by calculating tolerance granulations via deterministic operators or possible operators. Thirdly, If the intersection of the information granulations from the different attribute values is an empty set or isn't a subset of the elements from the new object sets, then it chooses the information granules of the attribute values according to different conditions. The execution is out of the loop until it doesn't satisfy the cycle conditions. It outputs the deterministic or possible decision rules. And lastly, it continues to find the deterministic or possible decision rules from the rest of the new object collections. The paper presents a new method that the deterministic or possible decision rules are mined based on granularity in incomplete information systems. It gives the mining algorithms and the instance. The approach has the advantages of high efficiency, more rules, concise forms, good comprehensibility and generalization ability.