Ensemble Approach to Sentiment Analysis of Google Play Store App Reviews

In the current digital era, sentiment analysis of Google Play Store application reviews has become a critical key to understanding public opinion on technology products. This study aims to evaluate the effectiveness of ensemble approaches in sentiment analysis compared to individual classification algorithms. The methods employed include ensemble techniques such as Random Forest and Boosting, along with individual algorithms like Naive Bayes and Support Vector Machine (SVM). This research incorporates extensive preprocessing steps, including cleaning, case folding, tokenization, stopword removal, and normalization, to prepare the data before classification. The results demonstrate that ensemble models, particularly Random Forest, achieve superior performance in sentiment classification of app reviews, with accuracy reaching 94.15% for Zoom app reviews and 80.69% for Shopee app reviews. This performance confirms that ensemble approaches are more effective in handling the complexity and variability of review data compared to individually operated algorithms. The study provides valuable insights for application developers to enhance their products based on user feedback. However, there is still room for improvement in terms of optimizing algorithms for highly unbalanced data and developing methods that can handle more complex language nuances. Recommendations for future research include the use of Deep Learning techniques and cross-domain testing to assess the effectiveness of these models in various sentiment analysis settings.