Leveraging Data-Driven Analytics for Mobile App Feature Extraction and Recommendations

Mobile app development necessitates extracting domain-specific, essential, and innovative features, aligning with user needs and market dynamics. However, identifying features to provide competitive edge to the app developers, is a non-trivial task often performed manually by product managers. This study addresses the challenge of mining and recommending app features by automatically identifying similar apps corresponding to the description of apps provided by the user. The proposed approach integrates Named Entity Recognition (NER) for feature extraction and BERT (Bidirectional Encoder Representations from Transformers) coupled with Topic Modeling for identifying similar apps. Our top-performing model, utilizing NMF for Topic Modeling with Sentence-BERT embeddings, achieves an F1 score of 87.38%, demonstrating its effectiveness in accurately identifying similar apps. Our contributions include compiling a dataset of 219 apps and 43800 user reviews to support research and development in feature recommendation. We have also developed an automated tool integrating NER for feature extraction and BERT-based similarity analysis. Through rigorous evaluation, we demonstrate significant performance improvements compared to existing solutions.