Research On Decision Tree Algorithm Based On Rough Set In The System Of Physical Health Examination

With the improvement of people's living standards, medical examination is becoming a new sunrise industry. In this paper, we have studied the technique of decision tree based on rough set used in the medical system. Research result shows that this technique is suitable to the medical auxiliary diagnosis system, in addition, having broad application background.Rough set theory is the mathematical analysis tool that used to deal with the uncertainty and incomplete information, focus on data-driven model, find out implicit knowledge, reveal the potential law, and establish a new way for data attributes analysis and knowledge discovery, which is increasingly used in the fields of data mining and intelligent decision. Meanwhile, another most important classification technology named decision tree is also very useful, which possesses the advantages of fast, efficient and easy to understand. comparing with other types of classification technology.From the perspective of the characteristics of medical domain, we propose a variable precision branch summarized roughness decision model by combining with the advantage between rough set theory and decision tree technique. Firstly, we put forward an improved domain reduction method on the basis of analyzing the attribute reduction algorithm of the current information system. Take weight as a criterion of measuring the importance of attributes, adding the important condition attributes gradually. In the meanwhile, delete the objects belonging to the domain U that can be completely correct classified in accordance with the subset of these attributes. Thereby, as the domain shrinking, it can also reflect the important properties of the decision system characteristics. And finally obtain a better reduction under the condition of maintaining the same classification accuracy.We also propose an access to decision rules on the basis of the theory of variable precision branch summarized roughness from the several accesses to decision rules combined with the features of the medical diagnosis, which solving the local optimal problems caused by a great number of access algorithm of decision rules based on rough set theory that only considering the comparison among the information inclusion degree, belonging to the current candidate attributes. In addition, make each branch contribution degree into the branch summarized roughness and regard it as the separation property selection criteria for the purpose of avoiding the local optimal effects, by evaluating the degree of each division of separation properties contributed to the whole classification. What is more, we introduce variable precision rough set model and error parameter |3 in order to solve the generalization problems better and improve the ability of classification, allowing to exist some classification errors when divide instances and make flexible adjustment between classification accuracy and generalization ability, which better adapted to the characteristics of medical diagnosis.