Research Of The Decision Tree In Data Mining Based On Semanteme

The traditional decision tree algorithm takes information gain as the rule to choose the attribute for classification. The attribute that has the biggest value of information gain can be selected firstly to build the decision tree. While calculating the value of information gain, the traditional decision tree algorithm does not include semantic information, it only simply considers the words'and characters'matching in grammar, it also lacks of the understanding of those semanteme information contained in the data. All of the above result in lack of intelligence and lead to heavy calculation, the complexity and the low-quality of classification and so on. Furthermore, the traditional decision tree algorithm will face a more big challenge for the large database.Based on the analysis of the decision tree algorithm and the corresponding concept such as HowNet, hierarchy tree, semanteme similarity and so on, this paper proposes a new decision tree algorithm based on semanteme. The new algorithm presents the method of separation of continuous-attributes and semantization of substantival-attributes, and builds up the semanteme-based decision tree model for data mining. The semanteme-based decision tree model can better use the semanteme information about the attributes in data sets; it can also meet the users'need of data mining based on the semanteme. To a certain extent, the semanteme-based decision tree model can achieve intelligent data mining.The results of the experiments show that the semanteme-based decision tree model can not only solve the problem of lacking of the semanteme information that exists in traditional decision tree algorithm, but also improve the expressing ability of knowledge in data mining systems. Comparing with the traditional decision tree mining system, the semanteme-based decision tree system has better efficiency and higher accuracy in the forecast.