Performance evaluation of the machine learning algorithms for emotion classification on the CASE dataset

Abstract

Emotion classification using physiological signals is still a challenging task even the sensor technology and machine learning algorithms evolved within the decades. In this study, the performance of KNN, DT, RF, LR, and XGB algorithms on emotion classification was evaluated in terms of accuracy on the CASE dataset. Three sub-datasets namely Downsampled, Resampled-EM, and Resampled-VA were obtained from the original dataset. Then, hyperparameter tuning was applied to the smallest dataset and the algorithms were applied with the parameters that were obtained in hyperparameter tuning to the Resampled-EM, Resampled-VA, and original sets. As the results obtained, KNN, RF, and XGB provided higher accuracies on the Resampled-VA set when compared to the Resampled-EM set, where it was the contrary for the DT algorithm. XGB algorithm provided the highest accuracy of 97.44% among all the results. This study could be considered as the most comprehensive study that utilizes machine learning algorithms for emotion classification on the CASE dataset.

Keywords

• Emotion recognition
• Machine learning
• Physiological signals
• CASE dataset

CASE veri seti üzerinde duygu sınıflandırma için makine öğrenmesi algoritmalarının performans değerlendirmesi

Öz

Sensör teknolojisi ve makine öğrenimi algoritmaları birkaç on yıl içinde evrimleşmiş olsa da fizyolojik sinyalleri kullanarak duygu sınıflandırması hala zorlu bir görevdir. Bu çalışmada, KNN, DT, RF, LR ve XGB algoritmalarının CASE veri seti üzerinde duygu sınıflandırmasındaki performansları değerlendirildi. Orijinal veri setinden Downsampled, Resampled-EM ve Resampled-VA olarak isimlendirilen 3 alt-veri seti elde edildi. Daha sonra, en küçük boyuta sahip veri setine hiperparametre ayarlaması uygulandı ve algoritmalar hiperparametre ayarlamasında elde edilen parametrelerle ResampledEM, Resampled-VA ve orijinal setlere uygulandı. Elde edilen sonuçlara göre, KNN, RF ve XGB algoritmaları Resampled-VA setinde DT algoritmasına kıyasla daha yüksek doğruluklar sağladı. Bu durum Resampled-EM seti için tam tersi olarak gözlemlendi. XGB algoritması, %97.44 ile tüm sonuçlar arasında en yüksek doğruluğu sağladı. Bu çalışma, CASE veri setinde duygu sınıflandırması için makine öğrenimi algoritmalarını en kapsamlı şekilde kullanan çalışma olarak değerlendirilebilir.

Anahtar Kelimeler

• Duygu tanıma
• Makine öğrenmesi
• Fizyolojik sinyaller
• CASE veri seti

Kaynakça

1. [1] Del Giudice M. “The Motivational Architecture of Emotions”. Editors: Al-Shawaf L, Shackelford TK. The Oxford Handbook of Evolution and the Emotions, 1-39, Oxford, UK, Oxford University Press, 2021.
2. [2] Alsharif AH, Salleh NZM, Baharun R. “The neural correlates of emotion in decision-making”. International Journal of Academic Research in Business and Social Sciences, 11(7), 64-77, 2021.
3. [3] Van Kleef GA, Côté S. “The social effects of emotions”. Annual review of psychology, 73(1), 629-658, 2022.
4. [4] Tammilehto J, Kuppens P, Bosmans G, Flykt M, Peltonen K, Vänskä M, Lindblom J. “Attachment orientation and dynamics of negative and positive emotions in daily life”. Journal of Research in Personality, 105, 1-12, 2023.
5. [5] Diener E, Thapa S, Tay L. “Positive emotions at work”. Annual Review of Organizational Psychology and Organizational Behavior, 7(1), 451-477, 2020.
6. [6] Mazzocco K, Masiero M, Carriero MC, Pravettoni G. “The role of emotions in cancer patients’ decision-making”. Ecancermedicalscience, 13(1), 914-936 2019.
7. [7] Keller A, Litzelman K, Wisk LE, Maddox T, Cheng ER, Creswell PD, Witt WP. “Does the perception that stress affects health matter? The association with health and mortality”. Health Psychology, 31(5), 677–684, 2012.
8. [8] Saxena A, Khanna A, Gupta D. “Emotion recognition and detection methods: a comprehensive survey”. Journal of Artificial Intelligence and Systems, 2(1), 53–79, 2020.

9. [9] Gouizi K, Bereksi RF, Maaoui C. Emotion recognition from physiological signals”. Journal of Medical Engineering & Technology, 35(6-7), 300-307, 2011.
10. [10] Pace-Schott EF, Amole MC, Aue T, Balconi M, Bylsma LM, Critchley H, Heath AD, Friedman BH, Gooding AEK, Gosseries O, Jovanovic T, Kirby LAJ, Kozlowska K, Laureys S, Lowe L, Magee K, Marin MF, Merner AR, Robinson JL, Smith RC, Spangler DP, Overveld MV, VanElzakker MB. “Physiological feelings”. Neuroscience & Biobehavioral Reviews, 103(1), 267-304, 2019.
11. [11] Zhang J, Yin Z, Chen P, Nichele S. “Emotion recognition using multi-modal data and machine learning techniques: A tutorial and review”. Information Fusion, 59(1), 103-126, 2020.
12. [12] Rim B, Sung NJ, Min S, Hong M. “Deep learning in physiological signal data: A survey”. Sensors, 20(4), 969-1008, 2020.
13. [13] Cui H, Liu A, Zhang X, Chen X, Wang K, Chen X. “EEG-based emotion recognition using an end-to-end regional-asymmetric convolutional neural network”. Knowledge-Based Systems, 205, 106243-106252, 2020.
14. [14] Hassan MM, Alam MGR, Uddin MZ, Huda S, Almogren A, Fortino G. “Human emotion recognition using deep belief network architecture”. Information Fusion, 51(1), 10-18, 2019.
15. [15] Hasnul MA, AbdulAziz NA, Abdulaziz A. “Augmenting ECG data with multiple filters for a better emotion recognition system”. Arabian Journal for Science and Engineering, 48(1), 1-22, 2023.
16. [16] Hssayeni MD, Ghoraani B. “Multi-modal physiological data fusion for affect estimation using deep learning”. IEEE Access, 9(1), 21642-21652, 2021.
17. [17] Bota P, Wang C, Fred A, Silva H. “Emotion assessment using feature fusion and decision fusion classification based on physiological data: Are we there yet?”. Sensors, 20(17), 4723-4740, 2020.
18. [18] Zhang T, El-Ali A, Wang C, Hanjalic A, Cesar P. “Corrnet: Fine-grained emotion recognition for video watching using wearable physiological sensors”. Sensors, 21(1), 52-77, 2020.
19. [19] Yıldız, E. R. & Bitirim, Y. “Performance Evaluation of KNN for Emotion and Valence-Arousal Classifications on CASE Dataset”. 12th International İstanbul Scientific Research Congress on Life, Engineering, and Applied Sciences, İstanbul, Türkiye, 21-23 January 2023.
20. [20] Sharma K, Castellini C, Van Den Broek EL, Albu-Schaeffer, A, Schwenker F. “A dataset of continuous affect annotations and physiological signals for emotion analysis”. Scientific Data, 6(1), 196-209, 2019.
21. [21] Russell JA. “A circumplex model of affect”. Journal of Personality and Social Psychology, 39(6), 1161-1179, 1980.
22. [22] Tübitak-Ulakbim. “Turkish National e-Science e-Infrastructure-TRUBA”. https://www.truba.gov.tr/index.php/en/main-page/ (08.02.2024)
23. [23] Scikit-Learn. “sklearn.model_selection.GridSearchCV”. https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html (08.02.2024).
24. [24] Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion, B, Grisel O, Blondel M, Prettenhofer P, Weiss R, Dubourg V, Vanderplas J, Passos A, Cournapean D, Brucher M, Perrot M, Duchesnay E. “Scikit-learn: Machine learning in Python”. The Journal of Machine Learning Research, 12(1), 2825-2830, 2011.
25. [25] DMLC XGBoost. “Python API Reference”. https://xgboost.readthedocs.io/en/stable/python/python_api.html (17.06.2023).