Sentiment Analysis of Public Opinion towards the Crypto Academy Class Using the Convolutional Neural Network Method

Abstract: Technological advances and the popularity of cryptocurrencies have encouraged many people to seek in-depth understanding through educational platforms such as Crypto Academy which is the largest cryptocurrency community in Indonesia. This research aims to analyze public sentiment towards Crypto Academy classes using the Convolutional Neural Network method. The dataset contains 3467 comments collected through scrapping from Youtube. The data goes through a pre-processing stage which includes cleansing, case folding, normalizing, tokenizing, stopwords, stemming and also TF-IDF weighting before being used in the CNN model. This research evaluates the performance of the model on two data sharing ratios of 80:20 and 70:30 and evaluates hyperparameters such as learning rate and optimizer. The results of the CNN model with optimal hyperparameters produced the best accuracy of 83.75% at a ratio of 70:30 with a learning rate of 0.0005 and an ADAM optimizer. Based on the results of sentiment analysis, it can be seen that sentiments tend to convey more positive sentiments with a total of 1580 positive comments compared to neutral sentiments totaling 1298 and negative sentiments totaling 548. This indicates that the majority of people's views on the Crypto Academy class are very good or positive.