Application of supervised machine learning models in human emotion classification using Tsallis entropy as a feature

Abstract Emotion identification acts as a critical component in passive brain-computer interfaces. The domain of EEG-based emotion identification has garnered substantial attention owing to advancements in machine learning models, notably in terms of higher accuracy and broader generalization capabi...

Full description

Saved in:
Bibliographic Details
Main Authors: Pragati Patel, Sivarenjani B., Ramesh Naidu Annavarapu
Format: Article
Language:English
Published: SpringerOpen 2025-05-01
Series:Journal of Big Data
Subjects:
Online Access:https://doi.org/10.1186/s40537-025-01177-8
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850268929367212032
author Pragati Patel
Sivarenjani B.
Ramesh Naidu Annavarapu
author_facet Pragati Patel
Sivarenjani B.
Ramesh Naidu Annavarapu
author_sort Pragati Patel
collection DOAJ
description Abstract Emotion identification acts as a critical component in passive brain-computer interfaces. The domain of EEG-based emotion identification has garnered substantial attention owing to advancements in machine learning models, notably in terms of higher accuracy and broader generalization capabilities. Current research investigates the utility of Tsallis entropy as an emotion recognition feature from EEG signals. For this study, different EEG rhythms are extracted via FIR filter and are normalized. Tsallis entropy is then computed using a sliding 4-s time window with a 2-s overlap, forming a feature vector for testing five machine learning models: K Nearest Neighbor (KNN), Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), Naïve Bayes (NB), and Decision Trees (DT). Additionally, the ensemble models KNN-DT and DT-LDA are also analyzed. The SEED dataset is employed for this study, and performance is evaluated through holdout cross-validation, considering accuracy, F1 score, precision, and recall metrices. Findings indicate that the DT-LDA ensemble model outperforms individual machine learning models and KNN-DT, achieving a significant accuracy increase of 5.6% to 11.2% for gamma rhythm. DT-LDA model displays the highest average classification accuracies 87.57% in gamma rhythm. This underscores the importance of high-frequency signals in emotion recognition, with gamma rhythm yielding the highest accuracy, while beta and all rhythms showing comparable results. These findings substantiate the efficacy of this methodology in the field of emotion identification utilizing EEG signals.
format Article
id doaj-art-27df0ed7461c443b835602c232e164c8
institution OA Journals
issn 2196-1115
language English
publishDate 2025-05-01
publisher SpringerOpen
record_format Article
series Journal of Big Data
spelling doaj-art-27df0ed7461c443b835602c232e164c82025-08-20T01:53:19ZengSpringerOpenJournal of Big Data2196-11152025-05-0112111810.1186/s40537-025-01177-8Application of supervised machine learning models in human emotion classification using Tsallis entropy as a featurePragati Patel0Sivarenjani B.1Ramesh Naidu Annavarapu2Department of Physics, Pondicherry UniversityDepartment of Physics, Pondicherry UniversityDepartment of Physics, Pondicherry UniversityAbstract Emotion identification acts as a critical component in passive brain-computer interfaces. The domain of EEG-based emotion identification has garnered substantial attention owing to advancements in machine learning models, notably in terms of higher accuracy and broader generalization capabilities. Current research investigates the utility of Tsallis entropy as an emotion recognition feature from EEG signals. For this study, different EEG rhythms are extracted via FIR filter and are normalized. Tsallis entropy is then computed using a sliding 4-s time window with a 2-s overlap, forming a feature vector for testing five machine learning models: K Nearest Neighbor (KNN), Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), Naïve Bayes (NB), and Decision Trees (DT). Additionally, the ensemble models KNN-DT and DT-LDA are also analyzed. The SEED dataset is employed for this study, and performance is evaluated through holdout cross-validation, considering accuracy, F1 score, precision, and recall metrices. Findings indicate that the DT-LDA ensemble model outperforms individual machine learning models and KNN-DT, achieving a significant accuracy increase of 5.6% to 11.2% for gamma rhythm. DT-LDA model displays the highest average classification accuracies 87.57% in gamma rhythm. This underscores the importance of high-frequency signals in emotion recognition, with gamma rhythm yielding the highest accuracy, while beta and all rhythms showing comparable results. These findings substantiate the efficacy of this methodology in the field of emotion identification utilizing EEG signals.https://doi.org/10.1186/s40537-025-01177-8Machine learning modelsEmotion recognitionTsallis entropyNonextensive statisticsEnsemble learning model
spellingShingle Pragati Patel
Sivarenjani B.
Ramesh Naidu Annavarapu
Application of supervised machine learning models in human emotion classification using Tsallis entropy as a feature
Journal of Big Data
Machine learning models
Emotion recognition
Tsallis entropy
Nonextensive statistics
Ensemble learning model
title Application of supervised machine learning models in human emotion classification using Tsallis entropy as a feature
title_full Application of supervised machine learning models in human emotion classification using Tsallis entropy as a feature
title_fullStr Application of supervised machine learning models in human emotion classification using Tsallis entropy as a feature
title_full_unstemmed Application of supervised machine learning models in human emotion classification using Tsallis entropy as a feature
title_short Application of supervised machine learning models in human emotion classification using Tsallis entropy as a feature
title_sort application of supervised machine learning models in human emotion classification using tsallis entropy as a feature
topic Machine learning models
Emotion recognition
Tsallis entropy
Nonextensive statistics
Ensemble learning model
url https://doi.org/10.1186/s40537-025-01177-8
work_keys_str_mv AT pragatipatel applicationofsupervisedmachinelearningmodelsinhumanemotionclassificationusingtsallisentropyasafeature
AT sivarenjanib applicationofsupervisedmachinelearningmodelsinhumanemotionclassificationusingtsallisentropyasafeature
AT rameshnaiduannavarapu applicationofsupervisedmachinelearningmodelsinhumanemotionclassificationusingtsallisentropyasafeature