The role of recommender systems is remarkable in the field of e-commerce. Multimedia contents – including live multimedia content – are gaining more and more significance on the internet. Due to the development of mobile devices the user generated multimedia contents gained more ground. Already existing methods for content recommending cannot be easily adapted for user generated live multimedia content. Both content-based and collaborative recommender systems have their disadvantages what make their utilization for live multimedia content recommending ineffective and inaccurate. My Thesis Design explains the characteristics and disadvantages of various content recommending methods. I describe the special features and problems of the planned system which come from the limited lifecycle of live multimedia contents and the system’s real-time operation. I write about the framework structure for testing and implementing the system and the co-operation of its parts: user behavior simulating-, the communication- and the recommender module. I describe how the lack of information in live multimedia content recomending can be solved when the explicit ratings of users are not available. I describe how the given problem can be solved using probabilistic matrix factorization with the help of the received informations, pointing out the revealed technical difficulties and their solutions. I show the process of the completed system’s testing and the data and results obtained during the tests and the conclusions that can be drawn. Finally, I propose and examine some of the possibilities for further development, which I couldn’t further test and examine because of the limitations of the simulated users.