<![CDATA[More than 3.5 million Indonesian educators had used the Merdeka Mengajar Platform by 2024. Comparing this figure to the 3.37 million from the previous academic year, there has been an increase of almost 3.85%. However, an investigation is required to determine the reasons why the application's use has not yet achieved the anticipated goal number of users. This study does sentiment analysis on evaluations of the Merdeka Mengajar platform using Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM). The benefits of RNN and LSTM in processing sequential data, especially in text processing for sentiment analysis, led to their selection. The purpose of this study is to solve the difficulties in determining if users' sentiments on the platform are favorable or negative. Important steps in the research technique include preprocessing, data cleaning, and employing FastText embedding to convert text into numerical vectors. Then, using patterns in the text data, RNN and LSTM models are used to forecast sentiment. The study's findings demonstrate that the LSTM model can, with an estimated accuracy of 93.58%, identify long-term associations in sequential data. The RNN model, on the other hand, produces a lesser accuracy of 91.70%. Particularly in text data with intricate temporal circumstances, the LSTM model performs better at accurately classifying sentiment. By better understanding customer opinions and input on the Merdeka Mengajar platform, this study helps platform developers improve the quality of their services.]]>
Copyrights © 2025
