Comparison of Automated Machine Learning (AutoML) Tools for Epileptic Seizure Detection Using Electroencephalograms (EEG)

Abstract

Epilepsy is a neurological disease characterized by recurrent seizures caused by abnormal electrical activity in the brain. One of the methods used to diagnose epilepsy is through electroencephalogram (EEG) analysis. EEG is a non-invasive medical test for quantifying electrical activity in the brain. Applying machine learning (ML) to EEG data for epilepsy diagnosis has the potential to be more accurate and efficient. However, expert knowledge is required to set up the ML model with correct hyperparameters. Automated machine learning (AutoML) tools aim to make ML more accessible to non-experts and automate many ML processes to create a high-performing ML model. This article explores the use of automated machine learning (AutoML) tools for diagnosing epilepsy using electroencephalogram (EEG) data. The study compares the performance of three different AutoML tools, AutoGluon, Auto-Sklearn, and Amazon Sagemaker, on three different datasets from the UC Irvine ML Repository, Bonn EEG time series dataset, and Zenodo. Performance measures used for evaluation include accuracy, F1 score, recall, and precision. The results show that all three AutoML tools were able to generate high-performing ML models for the diagnosis of epilepsy. The generated ML models perform better when the training dataset is larger in size. Amazon Sagemaker and Auto-Sklearn performed better with smaller datasets. This is the first study to compare several AutoML tools and shows that AutoML tools can be utilized to create well-performing solutions for the diagnosis of epilepsy via processing hard-to-analyze EEG timeseries data.

Time Series Forecasting Utilizing Automated Machine Learning (AutoML): A Comparative Analysis Study on Diverse Datasets

“…This paper evaluates multiple AutoML tools regarding how well they perform in analyzing time series datasets. The chosen tools were tested on metrics explained previously in the methodology, but there are other metrics that need examination as well [39]. These other metrics are not the scores given as to how accurately they are predicted.…”

Benchmarking Automated Machine Learning (AutoML) Frameworks for Object Detection

“…Borji et al produced a benchmarking study on the object detection of 41 different models and 7 datasets, but did not include AutoML frameworks [28]. Lenkala et al compared three different AutoML frameworks on three different time-series datasets for epileptic seizure detection [29]. A similar study of the aforementioned was performed by Westergaard et al, with the same AutoML frameworks and three new time-series datasets [30].…”

Advanced Dual RNN Architecture for Electrical Motor Fault Classification

“…In general, there are four main metrics used to evaluate ML classification models: accuracy, precision, recall, and F1 score [31]. Accuracy, as shown in Equation 13, is one of the most important metrics that represents the number of correct predictions (True Positives (TP) and True Negatives (TN)) divided by the total number of predictions (the sum of TP, TN, False Positives (FP), and False Negatives (FN)).…”