Identification of Cardiac Arrhythmia by Selection of Relevant Variables Using Genetic Algorithms

This article presents the computational identification of cardiac arrhythmia from electrocardiogram (ECG) signal recordings to facilitate timely diagnosis and clinical management. The cardiac arrhythmia dataset from the public UCI repository was used, comprising 279 features and 452 classified cases. Several variable selection algorithms were applied, including filter methods such as OneR, Chi-square, information gain, symmetric uncertainty, gain ratio, CFS, and consistency, as well as a metaheuristic approach based on the genetic algorithm. The variables identified through the filter methods were subsequently used as inputs for the OneR, PART, Rpart, JRip, C4.5, SVMlin, KNN, and random forest classifiers. The results indicate two subsets of particular interest: 37 relevant variables achieving an average balanced accuracy of 82.93% using the Random Forest classifier in combination with the CFS filter method, and 12 relevant variables yielding an average balanced accuracy of 82.56% when the Random Forest classifier is combined with the genetic algorithm. These outcomes were obtained without the application of any data balancing method. Smart citations: https://scite.ai/reports/10.61467/2007.1558.2026.v17i1.1145Dimensions.Open Alex.