Research On Automatic Recognition Of ECG Signals Based On Ensemble Learning

In recent years,cardiovascular disease has become one of the most important diseases that threaten human life,and the incidence is increasing year by year.Therefore,early and accurate diagnosis of ECG signal categories has important clinical medical significance.At present,some ECG signal recognition studies do not integrate various medically significant features,and some information hidden in the ECG is not easy to be found.In addition,a single model has performance limitations in solving the problem,which affects the prediction performance of the model,and the generalization ability of the model needs to be improved.To address the above problems,this paper applies ensemble learning to ECG signal classification and achieves good classification results.The main work is as follows:(1)Preprocess the ECG signal data and realize the basic classification model.First,the ECG data are preprocessed and locate the waveform by wavelet transform combined with threshold method.Then,features are extracted from three aspects:shape,interval,and amplitude.By analyzing the classification results of different feature subsets on the machine learning model,feature selection is completed and the best feature subset is obtained.Finally,the classification is implemented using the base model based on the best feature subset,each base model is trained to the optimum,and the results of the base model on the test set are compared.(2)An ECG signal recognition model based on Voting ensemble algorithm is proposed.Aiming at the problem that a single model has a performance limit during the learning process and is easy to overfit,a single-layer ensemble learning algorithm based on a voting mechanism is designed.The algorithm improves the shortcomings of a single algorithm by integrating several different base classifiers.The base classifier is determined according to the accuracy,variance,deviation,and interpretability of the model,and finally the classification results of each classifier are obtained by voting.The final experiment result show that the model is effective.(3)A model for ECG signal recognition based on an improved Stacking ensemble algorithm is proposed.In order to make full use of the value of the data and improve the generalization ability of the algorithm,this paper uses the Stacking ensemble framework to build a hybrid multilayer algorithm with two layers of classifiers,generating a dataset with the output of each base classifier,and then training a new algorithm on this dataset to obtain the final output.The algorithm is also improved to address the problem that the traditional Stacking ensemble algorithm ignores the differences between base classifiers,and gives different weights to the prediction results of the base classifiers according to their classification ability,which provides more valid input information to the secondary classifiers.The final experimental results show that the ECG signal classification model based on the optimal feature subset and the improved Stacking ensemble algorithm outperforms both the individual models and the traditional Stacking model in all aspects.