Comparative Analysis of Deep Neural Network Models for Time Series Classification

Abstract

The implementation of automatic Electrocargiogram (ECG) classification poses massive real world benefits, allowing a more streamlined patient diagnosis pathway for heart conditions that would not be found without cardiologist input. ECG Classification is a complex time series problem, and there are many machine learning solutions that can be used to analyse and classify ECG data - these are mostly hand-crated heuristics or feature selected. This paper poses five deep learning time series classification architectures that can classify ECG data swiftly and with increased performance than an average cardiologist, furthermore using Wilcoxon-Holm post-hoc analysis to determine if a set of classifers are statistically significantly different.

Kieran Molloy

Data Science Researcher

A highly competent data scientist involved in the Python and R open-source community - passionate about unsupervised learning techniques.