Research And Application Of Decision Tree Algorithm Based On Data Warehouse

Vocational training is an important means of human capital formation. The implementations of training in china often use unified approaches which ignore the philosophy of "individualized education" and have no relevance, so the overall quality of training is low. Applying Data Mining to the training field to find the potentially useful patterns and the hidden information, we can provide the basis of personalized training for the trainees and enhance the overall quality of training.Based on the research of classification techniques of traditional Data Mining, this thesis analyzes the features of training field and chooses the Decision Tree algorithm for the Data Mining of training field. We study ID3 algorithm deeply and proposes a new attribute selection algorithm to improve ID3 algorithm, with the help of some research results of domestic and foreign scholars. The new algorithm adds the number of the attribute's values and the class differentiation of the attribute to calculate the attribute information gain, and it overcomes the shortcomings of bias of multi-valued attribute and the selection of optimal properties of ID3 algorithm. Then, the effectiveness of the algorithm is verified through an instance and UCI data sets.Finally, in order to establish the model of decision tree, we apply the improved algorithm to the postal savings training system which is built with data warehouses and combined with the On Line Analytical Processing.