Abstract: A ensemble learning method (Ensemble Learning based on Denosing Autoencoder, ELDA) is proposed to study the accuracy of traditional classification learning algorithm. Unlike traditional ensemble learning approaches such as Bagging, Adaboost and Rotation Forest, ELDA first maps the data sets to a new feature space by using denoising autoencoder algorithm, then constructs the decision tree as the base classifier, and further classifies the data sets. Experimental results show that the accuracy of ELDA is higher than others, and it is proved that ELDA is an effective classifier ensemble algorithm of denosing autoencoder.