Optimized Sentiment Classification of Google Play Store App Ratings Using Advanced Machine Learning Models
Muhammad khan, Muhammad Qasim Khan, Fazal Malik, Nor Nabilah Syazana Abdul Rahman · VFAST Transactions on Software Engineering · 2024
The Google Play Store, the primary distribution platform for Android applications, hosts millions of apps and receives a large number of user reviews. However, extracting actionable insights from these reviews, particularly classifying rating sentiment (positive, negative, or neutral), remains a challenge. This paper addresses this issue by proposing a novel framework for app rating sentiment classification on Google Play Store data. We leverage a rich dataset of app reviews acquired from GitHub and employ a battery of advanced machine learning models. Specifically, we explore the efficacy of AdaBoost, XGBoost, and Artificial Neural Networks (ANNs) in conjunction with optimization techniques. Our approach significantly outperforms existing research, achieving superior accuracy ranging from 85% to 98% compared to the 78-95.9% accuracy reported in prior studies. This significant improvement translates to a deeper understanding of user sentiment across the app ecosystem. It enables developers to better gauge user satisfaction, prioritize improvements, and ultimately enhance user experience. Our work also paves the way for further research in sentiment analysis of app reviews, exploring fine-grained sentiment detection and analyzing sentiment dynamics over time.