Data Mining Based on the Rough Analysis and the Application in the Telecommunication Network Quality Evaluation

Abstract

The precision of the classification rule is decided by the construction of the classification algorithm. In this paper, after the introducing of the concepts and the attribute-reduction algorithms of the basic rough set, a data-mining algorithm based on the rough impend approximation measurement is Keywords: data mining, rough set, classification approximation, telecommunication network quality evaluationCopyright © 2016 Universitas Ahmad Dahlan. All rights reserved. IntroductionData Mining is concern with mining the valuable data from the massive data in the database to assist the decision-making. Presently, data mining is one of the research directions in the field of Information Decision and Artificial Intelligence. The Rough set method is a new information processing method [1,2], which is based on the classification of the equivalence relation. One of the characteristics of this method is, it is not needed to give the quantity description of some attributes in advance, rather finding out the internal rules of the problem directly from the description set of this problem [3,4]. It has great capability in analyzing the information system, which is incomplete, inaccuracy and includes noise.This Rough set is attended in the world because it has been successfully utilized in the field of Data Mining and Knowledge Discovering in database in recent years. There are three classification of attribute reduction based on rough set: (1) Attribute reduction algorithm based on discernibility matrix, reduction algorithm based on heuristic greedy, (2) Attribute reduction algorithm based on information indicated, and (3) Attribute reduction algorithm based on discernibility matrix was originally proposed by Skowron. In this algorithm, the problem of knowledge reduction of the decision table of information systems was converted to the problem of the reduction of discernibility matrix. Simple, easy to understand and achieve in practical application is the feature of the algorithm, while the disadvantage is that the algorithm reduction is inefficient and is a serious waste of space, not suitable for mass data.To address the above problem, various improved modifications of the difference matrix algorithm was got by reducing the time complexity of difference matrix reduction and many other ways by many scholars, making it a more efficient reduction algorithm. A property selection criteria was designed by greedy heuristic attribute reduction algorithm as the basis for selection of the current best attribute. Using heuristic ideas to reduce properties not only has the advantages in terms of time efficiency, and the final result is usually the best reduction or suboptimal, in fact, our needs in practical application can be satisfied by this result, so the heuristic algorithm is generally preferred as the attribute reduction.Rough set theory is generally indiscernible relationship-based, by introducing the approximation set and the lower approximation set, defined on the set operations, this is often