Scholarlink

Epileptic sseizure’s detection and classification is a challenging task as it happens for a short duration. Machine Learning techniques provide the way to enable faster and more accurate diagnosis and treatment. This paper investigates the use of ten mainly utilized ML models, including LR, NB, KNN, SGD, SVM, Random Forest, DT, ET, Gradient Boosting, and XGBoost, to evaluate and compare their performance in detecting and classifying seizures and non-seizures classes. This paper has used benchmark Bonn time series EEG dataset for seizure detection. The evaluation matrices F1-score, precision, recall, and accuracy were 100% for RF and NB. After RF and NB, the GB and XGB techniques shown F1-score, precision, recall, and accuracy as 97%, 98%, 97%, and 97.5% followed ET with 92%, 93%, 93%, 92.5%, and DT as 90% for each metrics, respectively. The findings indicate that Random Forest and Naïve Bayes demonstrated the highest performance, attaining a 100% accuracy score. The hyperparameter tuning and feature extraction impact analyzed on performance of learning technique for epilepsy classification shows good results. The machine learning with EEG signals shows great potential for detection and classification of seizures.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Epileptic Seizure Classification on Time-Series EEG Signals Using State-of-the-Art Learning Techniques

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Amit Kumar Yadav,
  • Bashir Alam,
  • Om Pal,
  • Jyoti Yadav

摘要

Epileptic sseizure’s detection and classification is a challenging task as it happens for a short duration. Machine Learning techniques provide the way to enable faster and more accurate diagnosis and treatment. This paper investigates the use of ten mainly utilized ML models, including LR, NB, KNN, SGD, SVM, Random Forest, DT, ET, Gradient Boosting, and XGBoost, to evaluate and compare their performance in detecting and classifying seizures and non-seizures classes. This paper has used benchmark Bonn time series EEG dataset for seizure detection. The evaluation matrices F1-score, precision, recall, and accuracy were 100% for RF and NB. After RF and NB, the GB and XGB techniques shown F1-score, precision, recall, and accuracy as 97%, 98%, 97%, and 97.5% followed ET with 92%, 93%, 93%, 92.5%, and DT as 90% for each metrics, respectively. The findings indicate that Random Forest and Naïve Bayes demonstrated the highest performance, attaining a 100% accuracy score. The hyperparameter tuning and feature extraction impact analyzed on performance of learning technique for epilepsy classification shows good results. The machine learning with EEG signals shows great potential for detection and classification of seizures.