Research On Interval Valued Decision Table Data Mining Based On Rough Set Theory

Data mining is now an important research branch of intelligent system. It combines advanced techniques such as artificial intelligence, computational intelligence(artificial neural network, genetic algorithm),pattern recognition and statistics to discover hidden valuable knowledge from databases.Rough Set Theory, introduced by Z.Pawlak in early 1980's, is a new mathematical tool to deal with vagueness and uncertainty. The basic idea is to derive classification rules of conception by knowledge reduction with the ability of unchanged classification. In recent years, rough set theory has become one of the most active research fields of artificial intelligence and information science, and has been successfully applied to many areas such as data mining, pattern recognition, machine learning, knowledge discovery, decision analysis, and so on.Firstly, the traditional rough sets based on the indiscernibility relation(also referred as equivalence relation) is introduced, which approximates sets of object by upper and lower set approximations. Secondly, as an extension of rough set, multigraded dominance-based rough set model is introduced this can deal with multi-attribute decision making problems with preference information instead of the traditional rough sets.Data mining in incomplete information system is a hard problem but inevitable in uncertain decision. In this thesis, an extended rough set model based on dominance relation is combined with fuzzy set theory for data mining in interval valued decision table, then decision rules can be obtained from the decision table. A ranking method in multi-attribute decision making with interval values is presented in this thesis. Decision rules are extracted from the decision system based on dominance rough set theory using a pairwise comparison table instead of the original one. Therefore, all alternatives are ranked and the best one is selected.