An Analysis of the Effect of Synaptic Weight Configuration for a Neural Network Enabled Handover for Heterogeneous Networks

Traditionally, Received Signal Strength (RSS) has been the primary indicator informing network selection strategies. However, approaches based on RSS are limited as they do not consider how (a) dynamic network conditions and (b) potential predictability of movement affects network performance. The wider research focus analyses the potential effect of weather on network handover decisions. In this context, a modulation strategy typically used in poor weather conditions is chosen and an analysis is done of the relative importance of the key dynamic performance metrics, loss, delay and RSS. In neural networks, synaptic weights reflect the relative importance of each performance metric. This work informs our selection of optimal synaptic weights when implementing a neural network controlled network handover decision within the context of the IEEE 802.21 Media Independent Handover (MIH) standard.