| Data mining is the fairly significant thesis in the field of information-processing technology, which consists of many theories and technology such as database,artificial intelligence,machine learning and statistics.Classification is one of the important functions of data mining,classification algorithm based on decision tree is widely used in data mining. Compared with other classification methods, decision tree has the following advantages: relatively smaller calculation workload, the ease of getting apparent rules, the cabability of showing important decision characteristics and higher correctness of classification, etc.However, the existing decision tree algorithm also exists a lot of shortage when being applied in practice,such as its lower computation efficiency and bigger scale of decision tree, etc. Therefore, it possesses significance both theoreticly and factually to make further improvement in decision tree algorithm so as to enhance its capacibility and make it more suitable for practical application.In order to try to solve the above problems, the author of this paper makes deep researeh on those points.The rough set theory is introduced into decision tree classification and the method of optimizing the decision tree classification algorithm is investigated. The main work done in this paper is as follows:Firstly, this paper introduces the related techonology and the theoretic basis of data mining and classification technology,and the emphasis is attached to the analysis and comparison of decision tree and post-pruning algorithms.Secondly, the decision tree algorithm is optimized in this paper in two aspects: attribute reduction and pruning. Attribute reduction algorithm ER based on the degree of dependency of attribute and post-pruning algorithm Prune based on rough set theory are proposed.Finally,the optimized decision tree algorithm is used in supplier measurement system,and its validity is verified when comparing with C4.5 algorithm. |
PDF Full Text Request