In this proposal, we study the problem of how to predict user application usage from our physical context, including context clues like location, time of day, and previously opened applications. The questions the thesis I am proposing will answer are: What context clues are important to predicting a user's behaviour? Can we summarize usage behaviour by a combination of context information? How can we use our context information to predict application usage? We explore some solutions in literature and propose our own solution, using Naïve Bayes networks and the Android framework to create an adaptive android application launcher that offers users quick access to relevant applications. We will explore the pairing of machine-learning algorithms and context-awareness, with the goal of providing a seamless intention → action experience.