基于机器学习的切换管理和模型浅析

Analysis of Machine learning Based Handover Management and Models

机器学习(ML)技术依赖于从蜂窝网络生成的数据。与其它优化技术相比具有优势,因为它可以学习难以解析出来的系统的隐藏模式和结构。由于蜂窝网络总是生成大量数据,例如信道矩阵,SNR和存储在信道状态信息(CSI)中的其它信息,因此获取数据很容易。ML可以利用这些已使用和未使用的数据来增强网络性能,包括切换优化问题。同样,基于ML的方法可以利用历史数据提前了解和预测网络参数,包括停留时间和基站(BS)流量,从而可以主动优化网络性能和用户QoS。本文介绍了超密集网络(UDN)中使用的基于ML的最新切换管理和切换模型。

Machine learning(ML)techniques rely on data generated from cellular networks.They have an advantage over analytical optimisation techniques because they can learn hidden patterns and structures in a system that is difficult to derive.Getting the data is easy since cellular networks always generate a massive amount of data such as channel matrix,SNR and other information stored in Channel State Information(CSI).Thus,ML can leverage these used and unused data to enhance network performance,including handover(HO)optimisation problem.Also,the ML-based approach can learn and predict network parameters including sojourn time,and BS traffic in advance by utilising historical data,thereby enabling proactive optimisation of network performances and user QoS.we present MLbased state-of-the-art HO management and HO models schemes that are used in UDN.