Komparasi Metode Machine Learning dan Deep Learning untuk Deteksi Emosi pada Text di Sosial Media

Abstract

Emotion Detection is the process of human emotions recognition, it extracting emotions such as happy, sad, and angry, which are obtained from human natural language. Linguistic Style has a wide range, emotional representations occur to millions of people and makes it difficult to infer a person's emotion in a concrete way. Multilabel datasets are also a challenge to deal in emotion detection. Therefore, an in-depth study of the appropriate method for emotional detection is needed. This study performs a comparative analysis between machine learning methods and deep learning methods. The machine learning methods used are Naïve Bayes, Random Forest, SVM, Gradient Boosting and Logistic Regression. The deep learning methods used in this study include LSTM, CNN, MLP, GRU and RNN. This research discovered that Deep learning has a better performance than machine learning, it seen from the accuracy values ​​of LSTM, CNN, MLP, GRU and RNN which exceed the accuracy values ​​of Naïve Bayes, SVM, Logistic Regression, Gradient Boosting and Random Forest.