Application of Convolutional Neural Network Algorithm for Analyzing Sentiments on the Kampus Merdeka Policy

Sentiment analysis examines public opinions on the Kampus Merdeka policy by analyzing texts from various sources. The study follows the Cross Industry Standard Process for Data Mining (CRISP-DM) method, encompassing stages such as business understanding, data understanding, data preprocessing, model implementation, and evaluation. The study utilizes preprocessing techniques, such as converting emoticons and emojis, text filtering, removing stopwords, stemming, word normalization, tokenization, and sequencing. The data for analysis is sourced from Twitter and YouTube, comprising 428 datasets. The accuracy, which measures the similarity between predicted and actual values, is 76%. Additional tests demonstrate that incorporating emoticon and emoji conversions in the text can increase sentiment analysis accuracy by 5%, resulting in 81%. These findings indicate the effectiveness of the Convolutional Neural Network algorithm employed in this research.