Non-Functional Requirements Analysis Based on Application Reviews in the Android App Market

Abstract

There are more than 3 million mobile apps in the Android market. The development process of every mobile application is rigorous, and many types of research on application quality requirements are derived, which are highly related to the development of applications. Research shows that user reviews of mobile applications are an unused large database that can provide feedback on user needs. In this article, user comments are automatically classified into non-functional requirements (NFRs) and other types. This paper proposes a loop matching classification technique (Loop Matching Classification). The three classification techniques of LMC, BOW, and TF-IDF were used to classify user comments, and the accuracy, recall rate, and F-measure of the results of the three classification techniques were compared. It was found that the Precision value of the LMC classification technique was 74.2%, the Recall was 82.5% and the F-measure was 78.1%.