Research On Decomposition Of Decision Information System Based On Rough Set Theory

Rough set theory is an approach of data analysis dealing with imprecise, inconsistent and incomplete information. The quantity and complexity of data in real-life application increase continuously and rough set data analysis method becomes inefficient. As an approach for data transformation, decomposition of decision information system is an effective way to improve the efficiency and quality of data analysis.The problems in decision information system analysis and the necessity of decomposition are discussed in this thesis. The concept of decomposition as well as its property and advantages are expatiated. The approaches of decision information system decomposition are divided into two types: the object set decomposition approach based on attribute significance can improve independence of classification and expansibility of rules, while the attribute set decomposition approach based on attribute dependency can reduce the complicated relationship between attributes of different sub-systems. Principles and algorithms of the two approaches are presented and analyzed in detail.The quality of data analysis and knowledge discovery depends on decomposition approach closely. Complexity of data analysis, accuracy of classification and equivalence of decision are proposed as criterions for evaluating the decomposition approaches. The equivalence of decision in attribute set decomposition is analyzed. Criterions of equivalence on the universe of decision information system and the sample space are proposed respectively to avoid the inconsistency of classification and the loss of information caused by decomposition.Experiments of the decomposition approaches are designed and implemented. UCI data sets are tested and the results are analyzed according to the criterions for evaluating the decomposition quality. The feasibility and availability of decision system decomposition in data analysis are interpreted experimentally.