Research On Attribute Reduction Methods Based On Information Entropy | Posted on:2005-05-23 | Degree:Master | Type:Thesis | Country:China | Candidate:Y Wu | Full Text:PDF | GTID:2178360185964152 | Subject:Computer software and theory | Abstract/Summary: | PDF Full Text Request | Attribute reduction is one of the important issues of rough set theory and is to remove superfluous knowledge from information systems while preserving the consistency of classfication. Attribute reduction is an important process in data mining based on rough set. Regarding the significance of attribute defined from the viewpoint of information theory as heuristic information, and introducing the heuristic information into genetic algorithm, an effective heuristic genetic algorithm for minimizing relative reduction is proposed. A modified heuristic algorithm of attribute reduction is presented. The construction of heuristic information is discussed in detail and the incompleteness of the two existing definitions of attribute significance is proved by two counterexamples. A modified definition of the attribute significance based on the weighted sum is proposed, and quantified it as heuristic information in reduction algorithm. The algorithm based on an viewpoint that knowledge is an ability of classing thing quantify knowledge and prove quantify reasonableness quantified capacity differentiate as heuristic information guiding reduce computation have improved reducing efficiency. Additionally using the heuristic information proposes a modified attribute reduction algorithm and modifies selection operator by knowledge magnitude,and it can achieve reductions as more as possible. Improving end condition to be determined by knowledge magnitude can achieve all the reduction for most of testing data sets, process relative reduction problems efficiently. | Keywords/Search Tags: | data mining, rough set, attribute reduction, entropy, knowledge magnitude, genetic algorithm | PDF Full Text Request | Related items |
| |
|