An Improved Decision Tree Classification Algorithm

This article based on the classification methods in data mining as the research object, compared classification method that is commonly used in data mining, thorough studied the decision tree classification model, analyzed and studied emphasis on the advantages and disadvantages of the ID3 algorithm and several existing improved algorithm in decision tree classification algorithm. Theoretically analyses the multivalued bias problem of ID3 algorithm, by drawing from the idea of OneR algorithm, came up with a information gain formula that based on error rate, Improved the information gain formula of ID3 algorithm, Solved the problem of ID3 algorithm multi-valued bias. On ID3 algorithm of logarithmic computation, computation efficiency is not well, this article adopts the method of approximation function to avoid a large number of logarithmic computation, improved the computational efficiency of the ID3 algorithm. In addition Business car customer database on the decision tree constructing experimental data sets for example,results show that the generated decision tree based on the error rate of information gain more conform to the objective reality,the number of leaf nodes is less,the decision tree is more concise. The classification experiments are carried out on the UCI database of four data sets. Experimental results show that the improved algorithm is much better than the ID3 algorithm, the accuracy of the constructed decision tree is improved,the computation efficiency is higher and the leaf of the constructed decision tree is less; data in the same number of samples the next set,the improved algorithm took less time to build a decision tree.