Robust Identification Algorithm of Network Communication Signals via Machine Learning Model

Abstract

The efficiency of communication processing and control depends heavily on the recognition of network signals, however irregularities and mistakes frequently arise during the application process. In this work, we leverage machine learning models to automatically identify computer network communication signals, leveraging recent advancements in artificial intelligence technology. In the simulation, we employed a support vector machine (SVM) model, and we utilized parameter optimization to address the overlearning issue. The process of classifying modulation signals involves the extraction of feature parameters through the application of support vector machine and radial basis function neural network (RBFNN) models, respectively. Real-world network communication involves the observation and collection of signals from various viewpoints or feature spaces. These views provide a variety of detailed insights into the signal, and feature extraction is carried out for each view to produce the associated feature vectors. An extensive description of the signal can be generated by extracting the features from several viewpoints. Various viewpoints' feature vectors are combined and synthesized. The robustness of signal recognition can be increased and the bias and inaccuracy that could be generated by a single view can be minimized by combining the data from several perspectives. The support vector machine performs better than the radial basis function neural network, according to experimental findings. When the signal-to-noise ratio (SNR) is high, network communication signals function effectively. However, the latter (RBFNN) performs significantly worse in low SNR settings whilst the former (SVM) retains good accuracy. Therefore, when it comes to computer network communication signals, the support vector machine model is thought to be more reliable.