Research and development of traffic prediction algorithms on cell level in the wide band cellular networks

Abstract

The paper performs research and development of more efective data traffic prediction algorithms in mobile networks using data clustering on the cell level. The developed algorithms were adapted to a specific problem, namely to data with shifted trend. For artificially generated data, that imit real network data, was applied Gausian mixture algorithm to separate the current trend, after last shifting, needed for a more accuare prediction with big timestamps. There was also estimated the potential impact of prediction error if not applying clustering methods.