Multi-label Feature Selection for Long-term Electrocardiogram Signals

Arrhythmia is a kind of cardiovascular disease that seriously threats human health, and intelligent analysis of electrocardiogram (ECG) is an effective method for the early prevention and precise treatment to arrhythmia. In clinical ECG databases, it is common to see the multi-label phenomenon that one patient would be labelled with multiple types of arrhythmia. However, the current research is mainly to use the multi-class methods for dealing with the problem of multi-label, ignoring the correlations between diseases and causing information loss. Therefore, this paper aims to propose a multi-label feature selection (MLFS) method based on ECG and design a novel evaluation criterion based on kernelized fuzzy rough sets so as to choose the optimal feature subset and optimize ECG feature space. Through sufficient experiments to prove the feasibility of our methods, we obtain the optimal feature subset composed of 23 ECG features. For the six evaluation criterions of multi-label learning, Average Precision is 0.8053, Hamming Loss is 0.1063, Ranking Loss is 0.1366, One-error is 0.2021, Coverage is 0.4018, and Micro-F1 is 0.5874. The outcome presents great superiority to the current algorithms of MLFS. This study is a prerequisite for implementing big data ECG classification diagnosis and disease modeling.