Convolutional Neural Network for Sentiment Analysis of Indonesian Online News
Brilliant Ayang Iswenda, Irhamah, Heri Kuswanto · 2024
The development of online news has made it the primary source of information for the public, replacing the previously dominant position of print media. In increasingly fierce competition, many media companies rely on page views to drive advertising revenue, making it one of their key performance indicators. By utilizing Convolutional Neural Networks (CNN), companies can extract sentiment from news headlines. The results of sentiment analysis predictions from online news headlines can be used to determine the proportion of popular news based on page views. This allows us to identify whether popular news (news with high page views) comes from negative or positive headlines. This information can be utilized by media outlets to create more engaging headlines to increase the number of popular news articles. By optimizing CNN parameters during training, the accuracy of the testing model can be enhanced. The results of this research show that sentiment analysis with Convolutional Neural Network shows the best accuracy result of 76.94% when using 100 filters and a filter size of 5. Positive news headlines are more likely to result in popular news compared to negative headlines