An ML-Based Quality Features Extraction (QFE) Framework for Android Apps

Context: The generic quality attributes fail to comprehend the current state-of-the-art challenges and constraints of mobile apps. Objectives: The goal of this study is to fill the gap in the systematic procedures to identify and extract specific quality features relevant to Android apps. Method: To accomplish the objective, we have proposed an ML-based Quality Features Extraction (QFE) framework for Android apps. QFE analyzes, parses, and gains insights from use reviews utilizing Natural Language Processing (NLP), Sentimental Analysis, Topic Modelling, and Lexical Semantics. Results: This study was tested on three different datasets and QFE successfully discovered 23 unique Android-specific quality features. Moreover, a comparative study with related studies was conducted and the analysis delineates that QFE provides a more reliable, efficient, and easy-to-use approach. Contribution: Briefly, (i) an ML-based empirical framework is proposed for discovering quality features for Android apps; (ii) the popular Topic Modelling technique is enhanced by RBLSALT, that is to automate the manual process of labeling topics in Topic Modelling; and finally, (iii) the pseudo-code and Python implemented notebook of the framework is also given to provide ease in the applicability of QFE. Conclusion and Future Work: Future work, is planned to evaluate the framework by comparing it with different techniques of feature extraction and to propose a specific features-oriented comprehensive quality model based on Android apps.