A QoS Equalization Algorithm Based on DAFDC-RNN Time Series and Multimodule TSK Fuzzy Systems for IoT Applications

Abstract To satisfy the multiple constraints required for link selection in the Internet of Things (IoT), a quality of service (QoS) link selection algorithm based on dual-stage attention, a full-dimension convolution based recurrent neural network (DAFDC-RNN) time series, and a multi-module Takagi-Sugeno-Kang (TSK) fuzzy systems is proposed. This load balancing system mainly solves the link overload problem for the new generation of IoT under the conditions of large-scale nodes and huge network traffic. First, the IoT resource-constrained scenario is abstracted as an NP-hard problem using an integer linear programming model. Short-term prediction of IoT traffic is then performed using a DAFDC-RNN. Finally, network resources in the multi-module TSK fuzzy system are used to select the most suitable routing links for different QoS conditions. The proposed method can solve the overloading problem for IoT links under QoS constraints without changing the physical system links.