Data Reduction And Rule Generation Based On Rough Set

Data mining, which is combination of database technique and artificial intelligence, is a promising research area, and at same time an important tool to improve decision support in management domain. Data mining is used to discover the implicit, previously unknown, and potentially useful information from a large amount of data.Rough set is a relatively new soft computing tool to deal with vagueness and uncertainty. It has received much attention of the researchers around the world. Rough set theory has been applied to many areas successfully including pattern recognition, machine learning, decision support, process control. In this paper rough set theory and the application of rough set theory to data reduction and mining rule is deeply researched. Attribute Reduction is one of the main topics in research on rough set theory. In this paper, we discuss the relationship between the condition attributes in terms of information entropy, and prove a useful inequality about relative reduction of attributes. A new significance of attributes in decision table is defined and is taken as a criterion to get a relative reduction. And then a heuristic algorithm based on that criterion is proposed. Finally experimental analysis of this algorithm shows it can obtain meaningful and small relative reduction.The vagueness of information about a decision may be caused by granularity of representation of information, so it is necessary to remove unnecessary granularity while keeping essential information. In this paper, DFRS an approach for data filtering based on rough set theory is investigated. We show the deficiencies of this approach and then improve it by merging values of more than one attribute. Finally an example is given to show our approach can not only effectively reduce granularity of information but also avoid mistakes DFRS can cause in some measure.Value reduction is rule reduction in essence. The traditional object of value reduction is getting the minimal reduction. This dissertation proposes the notion of support measurement, based on this notion we propose a new object of value reduction, that is getting the rules which have maximal support measurement. The new object dose not contradict traditional object and the data after value reductionwould keep original information in maximal degree. The relationship between rules is discussed, which could be used to facilitate searching value reduction. Based on the fore mentioned ideal an algorithm about value reduction is proposed, experiment show it is practical.Rule generation is an important technology in Data Mining. The instances after value reduction probably overfit the training examples, so they can not be used as decision rule. In the paper, we discuss the property of rule set. And then we improve value reduction method and based on Rough Theory a rule generation algorithm is presented. Then the parameter in this algorithm is analyzed. Experimental results show that this algorithm can effectively classify unknown data.