ECG signal feature selection method and system based on memetic algorithm

Abstract

The invention discloses a Memetic algorithm-based electrocardiography feature selection method and system. The Memetic algorithm-based electrocardiography feature selection method and system use an encapsulation-based feature selection framework and introduces a Memetic algorithm (Memetic Algorithm, MA) to perform optimal extraction on an electrocardiography. The method and system can extract features without positioning a basic waveform, so subsequent machine learning algorithm performance reduction due to waveform detection incorrectness can be avoided, and a variable actual electrocardiography can be effectively processed; and meanwhile, the calculated amount of waveform detection can be reduced, and the computation complexity is low. The method and system are independent of a preset feature index, require low level of prior knowledge, and can automatically extract feature information input in an original electrocardiography, in this way, irrationality due to an extraction index set by a man can be avoided. Compared with the conventional algorithms, the Memetic algorithm can effectively solve the problem of complex large-scale optimization, the Memetic algorithm can be used for optimization of selection vectors, and then a feature subset with representativeness can be obtained, and the prediction performance of subsequent classification / regression algorithm can be remarkably improved. In addition, sparse cost functions are added into adaptive value calculation, and the method can effectively avoid a locality problem and improve the generalization ability of the algorithm.

Description

technical field [0001] The invention relates to the field of electrocardiographic signal processing, in particular to a method and system for selecting electrocardiographic signal features based on a Memetic algorithm. Background technique [0002] Electrocardiography (ECG) is a time-varying potential signal generated by cardiac activity, which contains rich physiological state information and has important scientific research value and practical significance. It is widely used in medical research and clinical diagnosis, and can effectively detect various diseases related to the heart, and has gained more and more attention in recent years. Due to the high dimension of the ECG signal, it is generally necessary to use a feature extraction algorithm to reduce the dimension before analyzing it, so as to improve the accuracy and generalization ability of the subsequent machine learning process. In the prior art, two types of feature extraction algorithms are mainly used: (1) lo...

AI Technical Summary

Problems solved by technology

[0013] In view of the above-mentioned deficiencies in the prior art, the object of the present invention is to provide an electrocardiographic signal feature selection method and system based on the Memetic algorithm, aiming to solve the problems of poor learning effect, incomplete feature extraction, and low efficiency in existing feature extraction and selection methods. The problem of low accuracy

Images