Emotion is an unapproachable domain for researchers to understand the casual relationships and even mathematically prove the equation. However, it can be approached by using Electroencephalography (EEG) Test which gives the signals from part of the brain where it is correlated with emotion.
This paper proposes a simple Convolutional Neural Network (CNN) to classify emotional sentiment based on Electroencephalography (EEG) brainwave data. Our proposed model architecture has a structure of one dimensional CNN which is mainly used for signal data.
MLP model and Random Forest model, which respectively reached 94.89% and 97.89% accuracy with Information Gain method, are defined as baseline models. The experiment and evaluation results demonstrate the robustness of our proposed model which achieved over 97% test accuracy. With reduced dataset which is consisted of the top 2184 attributes applied by Information Gain method, the model achieved 98.13% test accuracy which outperforms both baseline models and our naive 1D-CNN model.