Research On Method Of Knowledge Acquisition Based On Rough Set Theory

Rough set theory proposed by Poland scholar Z.Pawlak in 1982 is a powerful tool for data analysis. It has gained increasingly studying in recent years, and has been successfully used in widely fields such as machine learning, decision analysis, process control, knowledge discovery in database, expert system etc. Therefore, studying the method of knowledge acquisition based on rough set theory has important significance and applied value.In this thesis, based on rough set theory we make an investigation on the methods of knowledge acquisition in information system. And the main results is as follows,In aspect of attribute reduction in decision table, this paper defines a kind of object set formed by partitioning objects with conditional attributes and decision attributes, studies this set, and proposes a heuristic algorithm for attribute reduction based on rough set and information entropy, by measuring the attribute significance and the relativity of attribute via combined the set with information entropy. This method of attribute reduction selects the optimum condition attribute by computing information entropy, gradually takes out the certain objects, reduces involved extension of objects, i.e. reduces elements studied, and develops the efficiency of reduction. And the algorithm for attribute reduction is applicable for both perfect decision table and imperfect decision table. These results have important instruction meaning towards studying relative reduction for decision table.In aspect of mining decision rules in information system, this paper measures the dependence of condition attribute on decision attribute based on the set proposed above, studied the rules mining method by the objects set in rough set. For decision table, this paper proposes an algorithm of dynamic mining decision rules based on partial granulation. This method is different from the single equivalence relation in the classical rough set theory; this algorithm can study one decision table from multi-angles and multi-hierarchies. The examples show that the algorithms are efficient. The abstract and specialization extent of the objects set can depict the ordinal partition of decision by condition attributes and their combination. By fining the granular at different angles and hierarchies, the algorithm can mine certain, non-redundant rules and can expand the method in imperfect decision table. These results shows the algorithm can provide favorable decision rules and can offer a new method made sense for the study of data mining.The paper has studied some methods of knowledge acquisition in information system based on rough set theory. The achieved results not only enrich and improve rough set theory, provide a new method for the study of data mining, but also are expected important applied value for data processing.