Recommending software features for mobile applications based on user interface comparison

App features are one of the most important factors that people consider when choosing apps. In order to satisfy users’ needs and attract their eyes, deciding what features should be added in next release becomes very important. Different from traditional requirement elimination, app stores provide a new platform for developers to gather requirements and perform market-wide analysis. Considering that software features provided to users can be found out by exploring existing apps, an important way to elicit requirements is analyzing existing features provided by products which offer related functions and then finding new trends and fashions promptly. In this context, we propose a data-driven approach for recommending software features of mobile applications based on user interface comparison. Our approach mines similar user interfaces (UIs) from publicly available online repository. To calculate UI similarity through the best matches of components of two UIs, text similarity is used to measure the similarity of UI components and genetic algorithm is introduced to improve the comparison efficiency. Then, we develop an algorithm to extract features from similar UIs based on a set of identification rules. These features are further clustered with text similarity algorithm and finally recommended to developers. The approach is empirically validated with 44 features from 10 UIs. The experiment results indicate that our recommended features are valuable for requirement elicitation. © 2018, Springer-Verlag London Ltd., part of Springer Nature.