Abstract
The purchase and download of new applications on all types of smartphones and tablet computers has become increasingly popular. On each mobile device, many applications are installed, often resulting in crowded icon-based interfaces. In this paper, we present a framework for the prediction of a user's future mobile application usage behavior. On the mobile device, the framework continuously monitors the user's previous use of applications together with several context parameters such as speed and location. Based on the retrieved information, the framework automatically deduces application usage patterns. These patterns define a correlation between a used application and the monitored context information or between different applications. Furthermore, by combining several context parameters, context profiles are automatically generated. These profiles typically match with real life situations such as ‘at home’ or ‘on the train’ and are used to delimit the number of possible patterns, increasing both the positive prediction rate and the scalability of the system. A concept demonstrator for Android OS was developed and the implemented algorithms were evaluated in a detailed simulation setup. It is shown that the developed algorithms perform very well with a true positive rate of up to 90% for the considered evaluation scenarios.