An ensemble learning framework for convolutional neural network based on multiple classifiers

Traditional machine learning methods have certain limitations in constructing high-precision estimation models and improving generalization ability, but ensemble learning that combines multiple different single models into one model is significantly better than that obtained by a single machine learning model. When the types of data sets are diversified and the scale is increasing, the ensemble learning algorithm has the problem of incomplete representation of features. At this time, convolutional neural network (CNN) with excellent feature learning ability makes up for the shortcomings of ensemble learning. In this paper, an ensemble learning framework for convolutional neural network based on multiple classifiers is proposed. First, this method mainly classifies UCI data sets using the ensemble learning algorithms based on multiple classifiers. Then, feature extraction is performed on the image data set MNIST using a convolutional neural network, and the extracted features are applied as input to be classified using an ensemble learning framework. The experimental results show that the accuracy of ensemble learning is higher than the accuracy of a single classifier and the accuracy of CNN + ensemble learning framework is higher than the accuracy of ensemble learning framework.