Optimizing Emotion Classification from EEG Signals: A Comparative Analysis of Optimization Techniques on the DEAP Dataset

Electroencephalogram (EEG) signals for emotion classification have gained significant attention in recent times due to its implicit applications in varied fields, including human-computer interaction, affective computing, and substantiated counseling sessions. This study set out to explore the effectiveness of convolutional neural networks (CNNs) in classifying emotions using the DEAP dataset, a publicly available repository of EEG recordings and corresponding emotional annotations by experimenting with different optimizers and loss functions, like “Adamax” and “binary_crossentropy,”, etc. By carefully manipulating these optimizers, the study aims to examine and determine the optimal combination with the highest classification accuracy. The findings contribute valuable insights into fine-tuning CNN models for emotion classification based on hyperparameters, paving the way for more robust and effective applications such as personalized diagnosis for counseling, stress-busting techniques and so on. The promising result of this research shows that the “Adamax” optimizer outperforms other optimizers by 82.98 % of the accuracy of classification.