The judgment of the type of arrhythmia is very important for the preven tion and treatment of cardiovascular diseases.In view of the shortcomings of multiscale fluctuation d ispersion entropy (FDE) in arrhythmia classification and recognition,such as single scale,unable to fully reflect the arrhythmia information and so on.By improving FDE features,An arrhythmia classification method based on adaptive multiscale fluctuation dispersion entropy (AMFDE) is proposed.Firstly,the signal is decom posed by using the variable mode decomposition (VMD) of the optimized K value before the FDE features are calculated,and then FDE of each scale subsequence is extracted as classification feature,and genetic algor ithm (GA) is used The support vector machine (SVM) penalty factor c and kernel funct ion parameter g are optimized,and finally the GA-SVM model is used for pattern recognition.Through the experimental verification,t he average accuracy,sensitivity and specificity of the proposed method for four types of heart rhythm are 95.3%, 95.4% and 98.4%,respectively. Compared with other methods,such as AMDE-SVM,it has obvious advantages,the pr oposed method has obvious advantages and can achieve accurate classification of electrocardiogram (ECG) signals.