|
Classification is an important task of data mining , decision tree model is commonly used data mining classification model has been proposed since 1966 has been widely studied and applied. However, since the decision tree construction process using a greedy algorithm , resulting in a decision tree is easy to over- fitting, size is too large , the resulting rule is too long and so on. To solve these problems , researchers have proposed a variety of optimization methods . In this paper, the existing decision tree methods were based on a comprehensive study , we propose a decision tree based on rough set theory optimization algorithm . The main work is as follows : ( 1 ) introduces the definition of KDD , the basic process , the scope and application of the main problems facing introduced the application of decision tree algorithm , commonly used classification models and the classical decision tree based classification model . ( 2 ) optimization of the existing types of decision tree technology introduced in detail , such as modifying the test attribute space , improve test attribute selection methods , decision tree pruning, data to be limited and changing data structures, etc. , and describes the each of these methods in more classic algorithms. While a variety of methods for qualitative analysis , comparing the advantages and disadvantages of each class method . ( 3 ) Based on the previous proposed multi- variable decision tree model inadequacies , this paper proposes an improved multi-variable decision tree algorithm VPMDT, the algorithm by selecting the rational combination of attributes as splitting attribute , can make the relative size of the tree small. By comparison experiments prove that our proposed algorithm is correct and effective. ( 4 ) Based on the above , implemented a prototype system that import data from a text constructed through the decision tree model , and can be calculated by importing test data classification model prediction accuracy . Through this system, able to analyze the performance difference between the comparison algorithm .
|