<![CDATA[PT. Digital Indonesia Berkat or what is called Tokocrypto is a cryptocurrency exchange company based in Jakarta that is officially registered with the Commodity Futures Trading Regulatory Agency (Bappebti). The level of customer satisfaction is of course very important to note so that companies can benefit and retain service users to continue using the company's services and attract potential service users. Public sentiment towards a cryptocurrency can also affect the price of the cryptocurrency such as the cryptocurrency owned by Tokocrypto called TKO. The amount of data provided by customers will take a long time to be analyzed manually. To overcome this, sentiment analysis can be carried out using a machine learning model that can understand the content of Tokocrypto customer feedback. This study applies the Bidirectional LSTM method to classify sentiment analysis using the tweet data of Tokocrypto service users. In addition, it is necessary to apply pre-processing of text data to overcome customer feedback which includes non-standard words and slang which causes the model to not understand the original meaning. The model is also adjusted for hyperparameters with the grid search method so that the model gets the optimal combination of parameters. Changing non-standard words does not guarantee increasing the accuracy of the model but can still help produce a better model with evaluation results including an f1-score value of 0.9485, a precision value of 0.9423, a recall value of 0.92, a training loss value of 0.0001 and a validation loss value of 0.0004.]]>
Copyrights © 2023
