Mining android applications for discovering API call usage patterns and trends

Abstract

Software libraries and frameworks, consisting of a collection of Class and Interface de nitions, provide a mechanism for code reuse by providing methods, APIs, components (generic functionality) and a support structure for developers to build applications, products and solutions. KitKat, Jelly Bean, Ice Cream Sandwich, Honeycomb and Gingerbread are di erent versions (open-source) of Android, one of the most popular mobile platforms in the world. In this thesis, we present the results of our large-scale (consisting of 1; 120 open-source applications and 17:4 million lines of code) API usage analysis of Android applications. Our work is motivated by the need to mine actual Android API usage, frequent API call usage patterns and trends to understand and generate empirical data on how developers are using the mobile platform in their applications. Extracting popular and frequently-invoked methods, API packages and API call-usage patterns is useful to both the API Producers and API Consumers. For example, API Producers can view the quantitative data on API usage as a feedback from users on the relevance, usability and applicability of the respective APIs. We conduct a series of experiments on analysing the Android platform API usage (usage of di erent packages, usage of methods, usage across categories) and present the results of our analysis using graphs such as Bubble Chart, Radar Chart, Heat-Map for effective visualization of the results and for extraction of actionable information.