Perbandingan Kinerja RNN dan CNN Dalam Klasifikasi Sentimen Ulasan Pengguna Aplikasi di Play Store

The public frequently shares their thoughts and opinions on various topics, such as products, public figures, or government policies, through online platforms. The process of analyzing review data is referred to as sentiment analysis. This study aims to compare the performance of two deep learning models Recurrent Neural Network (RNN) and Convolutional Neural Network (CNN) in classifying user sentiments across five review categories from the Google Play Store: design, photography, gaming, social media, and streaming. Choosing the right algorithm is essential to achieving optimal accuracy, given the variations in language and expression patterns within reviews. The dataset used in this study consists of 50,000 reviews with an imbalanced distribution of positive and negative sentiments. To address this imbalance, oversampling techniques were applied using the Synthetic Minority Oversampling Technique (SMOTE). The evaluation process measured each model's accuracy and loss levels. The results show that CNN consistently outperformed RNN across most categories. For the design category, CNN achieved the highest accuracy of 85% with a loss value of 0.41, compared to RNN, which achieved 83% accuracy and a loss of 0.53. On the other hand, the streaming category showed the lowest performance, with CNN achieving an accuracy of 69% and a loss of 0.63, while RNN achieved 67% accuracy with a loss of 0.72. These findings highlight CNN's superior effectiveness in sentiment analysis across diverse user review categories.