The Impact of Smart Home App Review Characteristics on Review Helpfulness: A Topic Modeling and Machine Learning Approach
Yeji Kim, Kyuhong Park, Dongyeon Kim · The Institute of Management and Economy Research · 2025
Purpose - This study investigates the determinants of review helpfulness for smart home apps by examining latent topics within user reviews. It aims to discern how general issues versus specific product integration concerns influence perceived review utility. Design/methodology/approach - Smart home app reviews were collected from the Google Play Store for applications including LG ThinQ, Samsung SmartThings, and Google Home. Following rigorous preprocessing, the BERTopic framework was applied to extract five latent topics. These topic probabilities, alongside review characteristics, were incorporated into regression and machine learning models to predict review helpfulness. Feature importance analysis was conducted to evaluate the contribution of each topic. Findings - Both regression and machine learning results indicate that reviews addressing broad issues, such as login errors and application device integration, are more likely to be deemed helpful. In contrast, topics centered on specific IoT product integrations (e.g., refrigerators, air purifiers, washing machines, dryers) tend to decrease perceived helpfulness. Research implications or Originality - By integrating advanced topic modeling with predictive analytics, this study offers novel insights into the smart home domain. The findings provide actionable guidelines for enhancing app usability and optimizing review presentation, thereby contributing to both academic research and industry practices.