Multi-access edge computing (MEC) can support context-aware and delay-sensitive applications and allow cloud service computation to run at the edge of the network. MEC can utilize a wireless network to connect all kinds of computation-constrained devices to rapidly analyze data in a real-time manner. However, the explosion of these connected devices to deliver data will result in low network capacity and high network latency due to the uncertainty of devices choosing the right MEC server. In this paper, we define multi-server MEC wireless networks (MMWNs) as a type of wireless communication network where each device in the network can communicate with an MEC server directly. This paper explores using the broad learning system (BLS) to optimize network performance in MMWNs. Besides that, we introduce two different network allocation strategies based on the BLS scheme. Through simulations, our results reveal that the proposed strategies can significantly outperform the original BLS scheme in terms of network capacity and network latency.