The objective of my master’s thesis design is improving and testing a simulator created formerly and based on a user behaviour model, for giving simulated user inputs for an experimental live-video recommender system.
In this thesis I introduce references about user behaviour on video content sharing portals, results of measurements on existing live-video systems, and I describe a statistical model of user behavior based on these results considering the specialties of the mobile environment. I give an overview of a whole live-video system that contains a content recommender module and previously implemented parts of a simulator module that generates user activities to evaluate the recommender.
In the second part I present the results of my work, such as improving the existing parts of the user behavior model and extending it with new parts to satisfy the requirements of the live-video portal, developing the social tagging method, adjusting the parameter values used in the model towards the expected results, and implementing the model’s modifications. I report various debugging and testing methods and test results of the simulator and its specific parts as well, and I suggest some additional kinds of test.
Finally I summarize the results of my thesis work and I mention the practical uses and the possibilities for improving the user simulator.