Heuristic-Rule Based Model for Packet Loss Inference in IIoT Networks

IIoT Networks' performance is significantly impacted by packet losses in the network, which is the failure of data packets in reaching their intended destination within the network. Most of the transmission control protocol versions reduce the rate of transmission during the detection of packet losses, assuming network congestion and interference, thus resulting in operational disruption, reduced efficiency, data integrity failure and economic impact. However, not all packet losses are due to congestions and interference, some happen based on link issues from wireless which are seen as non-congestive packet losses as most transmission control protocol (TCP) modifications reduce the rate of transmission when these losses are detected while assuming network congestion, so TCP could not at present distinguish among these types of packet losses and reduces the rate of transmission irrespective of the types thus resulting in lower throughput for IIoT networks clients.
In addressing this issue, a heuristic-rule-based machine learning model was used for packet loss identification, classification, and prediction to differentiate between the types of packet losses at the IIoT network hosts’ end. The result shows that Random Forest performs better based on the rule, giving a hopeful resolution to an enhanced IIoT network's performance.