Analysis of Arrhythmia Features based on LightGBM and Kmeans Feature Extraction

Abstract

The risk classification based on heart disease signals aims to classify patients into different risk levels, such as low risk, medium risk, and high risk. This classification can help doctors determine appropriate treatment plans, take timely measures, and improve patients' survival rate and quality of life. This article is based on the background of this problem, extracting electrocardiogram signals, and using LightGBM and Kmeans algorithms for arrhythmia classification and analysis. Extracted from the 2s features collected in this article are the data features of the electrocardiogram, including P-wave, P-R band, QRS complex, S-T band, and T-wave features. Construct a feature index pool for each electrocardiogram data, and use the feature index pool and the assigned risk level to construct a LightGBM classification regression model. The prediction accuracy is 93.5%, and the error fluctuation does not exceed ± 1. After that, use principal component analysis (PCA) to perform feature dimensionality reduction on the data. Specific classification of different categories of heart rates was carried out, and the specific performance of the algorithm and visualization of classification images were performed. Finally, this article conducted model validation on the above model, and constructed training and validation sets to validate the model. Through observation, it can be seen that the accuracy of the model is stable at over 90%. Therefore, the model constructed in this article is true, effective, reliable, and accurate.