Emotion Classification in Indonesian Language: A CNN Approach with Hyperband Tuning

Abstract

In today's world, there is a high demand for accurate techniques to classify emotions in various fields. This study proposed utilizing a Convolutional Neural Network (CNN) optimized with a Hyperband Tuner (HT) to perform the Emotion Classification task in the Indonesian language effectively. Various feature extraction techniques experiments were conducted to explore the best combinations of feature extraction and CNN for the data set, including CountVectorizer (CV), TF-IDF, and Keras Tokenizer (KT). Last, the proposed methodology was evaluated and compared to the stateof-the-art techniques, including K-Nearest Neighbors (KNN), Decision Tree (DT), Naive Bayes (NB), and Boosting SVM. The experimental results revealed that the proposed method in this research outperforms the existing technique as evidenced by the accuracy, precision, recall, and F1-score metrics, which respectively reached 71.5655%, 71.5483%, 71.5655%, and 71.0041%.