A Method Based on Granule to Mine Classification Rules

Abstract

This paper makes use of knowledge granular to present a new method to mine rules based on granule. First, use the measure to measure the importance of attribute, and get the granularity of the universe, and then repeat this procedure to every granule of the granularity, until the decision attribute has only one value for all granules, then we will describe every granule to get the rule. The analysis of the algorithm and the experiment show that the method presented is effective and reliable. Classification rules are the main target of association rule, decision tree and rough sets. A new algorithm to mine classification rules based on the importance of attribute value supported. This algorithm views the importance as the number of tuple pair that can be discernible by it, and the rules obtained from the constructed decision tree is equivalent to those obtained from ID3, which can be proved by the idea of rule fusion. However, this method is of low computation, and is more suitable to large database. Rough sets is a techniques applied to data mining problems. This paper presents a new method to extract efficiently classification rules from decision table. The new model uses rough set theory to help in decreasing the calculation need for building decision tree by using what is called reduct algorithm and a rules set (knowledge) is generated from the decision table. The reliable classifier architecture is obtained and its effectiveness is verified by the experiments comparing with traditional rough set approaches. Data mining research has made much effort to apply various mining algorithms efficiently on large databases.