<![CDATA[Mutual fund investment is increasing, as evidenced by the increasing number of mutual fund application users on the Playstore platform in Indonesia. The Financial Services Authority (OJK) reported that the number of mutual funds in Indonesia until August 2022 reached 2,193 units. In this research, the data collection used is the data scrapping method on the Google Playstore website. The result of the scrapping data is an excel-formatted document of 3000 data which is then stored and processed using the Long Short Term Memory (LSTM) model. In order to facilitate the modeling stage later, the sentiment review data must go through a text preprocessing process. To improve the performance and performance of LSTM modeling more optimally, then in this study a choice of hyperparameters was made. The hyperparameters tested are Epoch, Batch Size and Layer LSTM. The highest accuracy value on the Ajaib dataset is 99.3% which is located at epoch 32 and batch size 50, the highest accuracy value on the Bareksa dataset is 95.1% which is located at epoch 32 and batch size 50, and the highest accuracy value on the Bibit dataset is 94.9% which is located at epoch and batch size 50. So that the highest accuracy value among the three datasets is obtained by the Ajaib dataset where the accuracy reaches 99.3%. From the test results of the three parameters, it proves that there is an increase in accuracy results that is good enough to reach the highest accuracy value of 0.9933.]]>
