Hybrid AI Framework for High-Precision Signal Quality Classification in Optical and Wireless Communication Systems

Abstract

Ensuring reliable signal transmission has become a critical requirement in modern optical and wireless communication systems, particularly as data networks evolve toward higher bandwidth, lower latency, and improved noise resilience. Traditional signal assessment techniques often rely on manual thresholding or single-model learning approaches, which struggle to perform effectively in noisy environments, imbalanced datasets, and complex multi-feature interactions. This limitation creates significant challenges in accurately predicting signal quality across diverse modulation formats, receiver parameters, and varying channel conditions. To address these challenges, the proposed system introduces a data-driven, automated, and hybrid artificial intelligence pipeline designed to classify signal integrity with high precision. The workflow incorporates comprehensive data preprocessing techniques, including label encoding, missing value handling, and Synthetic Minority Over-Sampling Technique (SMOTE) for class balancing to mitigate skewed data distributions. The system evaluates multiple Machine Learning (ML) models, including Complement Naive Bayes (CNB), Support Vector Machines (SVM), Linear Discriminant Analysis (LDA), and a proposed Dual-Learner Fusion Architecture (DLFA). The DLFA combines Logistic Regression (LR) and Random Forest (RF) using a Voting Classifier to leverage the probabilistic strengths of LR and the non-linear feature learning capabilities of RF. Model performance is evaluated using advanced metrics such as accuracy, confusion matrices, and ROC curves. To enhance usability, a fully interactive Tkinter-based Graphical User Interface (GUI) is developed, enabling dataset uploads, preprocessing, Exploratory Data analysis (EDA) visualization, model evaluation, and real-time signal quality prediction. Experimental results demonstrate that the DLFA model improves decision boundary learning, resulting in enhanced robustness, improved generalization, and greater interpretability for signal integrity assessment in optical communication systems