Wireless Network Interferences Prediction and MAC-Layer Optimization Using QNN

Abstract

Millions of devices are using wireless communication, which is expanding 
quickly. These devices can interact with one another in a dynamic environment, 
and providing uninterrupted services is more crucial. Conventional machine 
learning techniques were utilized to deliver good services, but they have 
significant drawbacks, such as their inability to handle correlated signals, their 
reliance on probability theory, and their inability to handle nonlinear signals. In 
order to improve MAC layer decision prediction accuracy and give users 
uninterrupted services, we implemented QNN (Quantum Neural Network) 
based wireless communication in this study. QNN is able to handle multiple 
network interactions at a time with limited features. In this model we used three 
qubits with two hundred layers variable circuit, dataset opted with more real
time wireless interactions with high SNR values, abnormal behavior, a greater 
number of retry requests, and with heave inferences. Experiments are carried 
out to evaluate performance of QNN approach, the results are also compared 
with TNN, LSTM, RF, XGBoost, and SVM are evaluated are evaluated in terms 
of MSE, MAE, RMSE, R2, PDR, Connection probability, Collision Reduction, 
variance, Cosine Similarity, and Euclidian error. QNN exhibits low MSE value, 
low MAE value, low RMSE value, high PDR value, high R2 value, high 
connection probability, high collision reduction value, low variance, high 
cosine similarity, and low Euclidian error when compared with other similar 
approaches that are used for prediction purpose.