Improved C4.5 decision tree algorithm based on sample selection

To improve the classification accuracy and reduce the training time of large sample, and find the best training set, this paper proposes the improved C4.5 decision tree algorithm based on sample selection. The algorithm is based on the fact that decision tree can only get local optimal solution and has the bigger relativity with initial sample. In sample selection, we use iteration process to find the best training set. Using accuracy of the selected sample training as iteration Information is highly optimized for general use. Partition similarity is used for the best selection as the standard. Experiments show that the accuracy and time consumption of the proposed algorithm, aiming at classification and identification of large sample data, is better than C4.5 decision tree.