Research On The Attribute Reduction And Classification Algorithm Of The Probability Value Fuzzy Decision System Based On Rough Set

With the rapid development of Internet technology,people have more and more data,and the types of data are more and more rich.As well known,symbolic data and numerical data are two commonly used types of data.In symbolic data,each object can only take one value.Real-valued data can take any value in the domain,there is no discrete concept.However,if the data representa-tion is in between the two kinds of data presentation,for example,for each attribute,it is no longer to take only single value,but each attribute value has a certain possibility.At this time,the two data formats have not met the needs of the people.Therefore,people put forward a fuzzy informa-tion system.In the fuzzy information system,each attribute value is no longer a value,but a fuzzy set,indicating the degree to which the attribute value belongs.In traditional discrete information systems,attribute values under the same attribute are mutually exclusive.However,each attribute value of the fuzzy set information system is only related to the object,and the mutually exclusive relationship between the attribute values under the same attribute is not disclosed.Based on the fuzzy information system,this paper proposes a probability value fuzzy decision system,that is,for each object,it forms a probability distribution on one attribute.In this case,each attribute is not only related to the object,but also influenced and restricted by other attribute values under the same attribute,so as to discover the mutually exclusive relationship between the attribute values under the same attribute.Probability is widely used in people's study and life.The characteristics of probability make the probability value fuzzy decision system satisfy many properties.Rough set theory is a mathematic instrument proposed by Z.Pawlak,a Polish scholar,to effectively de-scribe the uncertainty.At present,rough set theory has been widely used in many fields,of which attribute reduction is one of the most important applications.In this paper,the probabiility value fuzzy decision system is combined with the rough set theory.Under the framework of the rough set,three kinds of upper and lower approximation concepts are defined and their relations are stud-ied.Based on this,three attribute reduction algorithms and classification algorithms based on fuzzy decision tree are proposed,and the effectiveness of the algorithm is verified through experiments.Specifically,the main work of this paper is as follows:·Based on KL divergence,a measure of similarity between attributes is defined,·Three lower and upper approximation operators are proposed and the relationship between the three models is studied.The concepts of accuracy,roughness and approximate accuracy are defined,and their monotony is proved,·Based on the probability value fuzzy decision system,a new conditional entropy formula is proposed and its monotonicity is proved.Under new conditions Entropy and three ap-proximate operators,three attribute reduction algorithms are proposed,which are verified by experiments and proved the effectiveness of the algorithm,·Proposed three new algorithms for constructing fuzzy decision trees.Compared with the existing algorithms,it is proved that the algorithm is valid.