ABSTRACT
COVID-19 pandemic makes students can ony continue their education through the E-Learning system. In order to fulfill the goal of E-Learning, learning center departments of educational institutes need to know what the user needs and the only way to communicate with them is through feedback. However, it is a hard and time-consuming task to extract value from a large amount of feedback. This paper aims to implement and evaluate several RNN architectures’ performances including Simple RNN, Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU), to be able to classify feedback text to its categories via a multiclass classification approach. Furthermore, this paper uses FastText in comparison with Keras Embedding Layer to extract features along with the use of Random Oversampling and SMOTE in order to deal with imbalanced dataset problem. Based on the result, our final model could achieve a macro-averaged F1-Score of 64.35% using LSTM architectures. Furthermore, our paper shows that FastText has a poor performance in every RNN architectures and Random Oversampling has a better performance than SMOTE in handling imbalanced dataset problem. © 2020, World Academy of Research in Science and Engineering. All rights reserved.