Fig. 8From: Machine learning-based dynamic frequency and bandwidth allocation in self-organized LTE dense small cell deploymentsRandom sampling with 10 % of 337 permutations being explored with the linear and the bagging tree regression methods and the aggregate regressor. a, b TCP and proportional fair scheduler. c, d UDP and round robin scheduler. a Normalized performance considering per user optimization. b RMSE over user performance regression fitting with actual measured performance ranging from 538 to 1545 kbps. c Normalized performance considering per user optimization. d RMSE over user performance regression fitting with actual measured performance ranging from 1521 to 2871 kbpsBack to article page