Research Of Attribute Reduction Base On Two Rough Set Models

As an important method of intelligent information processing,rough set model was firstly proposed by the Poland scientist Z.Pawlak.It becomes very popular and also attracts many scholars.Many optimized rough set models have been proposed by them for breaking the limitations of rough set application during research.It is worth mentioning that fuzzy rough set model and decision-theoretic rough set are two of the most popular rough set models.The former is used to overcome the weakness of rough set theory in dealing with fuzzy problems,so that rough set theory has a better processing ability in dealing with fuzzy problems.The latter is a rough set model which is improved by Bayesian risk decision theory.It introduces semantic interpretation,which alleviates the awkward situation of rough set solving problems lacking scientific semantic support.At the same time,it also eliminates the contradiction between error rate and zero error tolerance rate of classical rough set.With the popularity of these two rough set models,the research on attribute reduction which is one of the most important concepts of rough set theory are becoming more and more valuable.In this thesis,we will analyze from the aspects of models and applications,the main content in the thesis are as below:(1)Aiming at fuzzy rough set model,we introduce the theoretical knowledge of the model itself,and introduced from the aspect of test cost.Then,we analyze method of test cost sensitive of attribute reduction,and give two different algorithms with core concepts of them.At last,we carry out experiments of these concepts to verify out theory.From the experimental results,we can find that the genetic algorithm can get more efficient results without considering the time.(2)Aiming at the decision-theoretic rough set model(DTRS),we study the method of attribute reductions of decision-preservation and decision-monotinicity based on decision rule from the aspect of decision rule.And then,we introduce the idea of local into DTRS model,and give the definitions of local decision-preservation attribute reduction and local decision-monotinicity attribute reduction.In the next step,we analyze the feasibility in theory.At last,we carry out experiments on the idea of algorithm given out before to verify the feasibility of the local attribute reduction.From the experiment result we can find that Local decision preservation and decision monotonic reduction algorithms have better processing ability in two aspects:reducing redundant attributes and obtaining decision rules.