Research On Knowledge Acquisition From Partial Order Decision Table Based On Rough Sets Theory

Rough set theory, proposed by Z.Pawlak in the early 1980s, is mathematical theory for reasoning about data Attributes reduction, the core of the Rough Set Theory, which is a new type of theory as the tool dealing with the uncertain knowledge, is the focal point of algorithm research for rough set. Its basic idea is to derive classification rules of conception by knowledge reduction with the ability of classification unchanged. Compared with the traditional method of uncertain data processes, the most significant feature of Rough theory is that it doesn't need any experience information of data Under the original data, it bases on equivalence relation to directly classify universe, and use the concept of upper and lower approximation to describe object.In the paper, an overview of the current situation of researches on Rough Set is detailed at first. Characteristic and deficiency of Rough set theory are analyzed in depth. And then, we introduce the foundation of Rough set theory, general attribute reduction algorithms and heuristic attribute reduction algorithms based on core are introduced. As is known to all, in classical Rough Set theory, equivalence relationship on relation domain plays a vital role. But in reality, the dual relationship on relation domain is not equivalent to the regular, the application of classical rough set model will be limited, for example, there is a partial order issue in the dual relationship, we can't create equivalence relations through attribute, In such circumstances, how to import partial order to the Rough set theory has been becoming the foundation, though which, we can study the attribute reduction algorithms, core computation in depth.This dissertation is mainly focused on the following aspects.1)Analyze and summarize the current situation of researches on Rough Set systematically, based on the analysis to characteristic and deficiency of Rough set theory, look forward to the development prospects of Rough Set theory.2)The all-attributes reduction and the minimum attributes reduction in classical rough set theory are all NP-hard problem, the major reason is that the minimum attributes reduction must try various combinations of attributes. This paper elaborates the classical rough set model, the basic concept of decision table, discemibility matrix and attribute reduction, introduces general attribute reduction algorithms and heuristic attribute reduction algorithms based on core.3) In this paper, we sort all the objects by each attribute values in decision table, and mine the rules of overall ranking. For this reason, we import partial order relation(reflexivity, anti-reflexivity, transitivity) into decision table, and we get the partial order decision table. In this foundation, we do data analysis and simplify decision rule.4) In the foundation of partial order decision table, we analyze equivalence classes in partial order decision table, expatiates methods of core and attribute reduction from those classes, finally, an example illustrates the efficiency of those methods.