A Hybrid Active Learning and Machine Learning Model for Intelligent Decision-Making in Wireless Communication Systems

Abstract

Efficient decision-making plays a vital role in the development of more efficient and reliable wireless communication systems in dynamic conditions. Existing machine learning methods are primarily based on decision precision and utilize certain forms of uncertainty-based sampling or retraining, adding a higher cost associated with labeling and a lack of decision-making efficiency. The study proposes a novel performance-centric HAL-ID (Hybrid Active Learning-Based Intelligent Decision) model integrating hybrid active learning and an optimization utility-based decision model. In addition, three unique components are included in the proposed method, such as a dual criteria-based sampling procedure incorporating predictive entropy learning with estimated wireless utility gains, adaptive retraining based on performance degradation indicators, and a decision-making rule based on a utility optimization technique. It used UrbanMIMOMap and Wireless-Intelligence datasets, achieving a decision accuracy of 95.8%, a spectral efficiency of 6.74bps/Hz, a packet error rate of 3.9%, and a labeling cost reduction of 54-58%. The suggested intelligent decision-making model based on hybrid active learning combines entropy-utility-based sampling, utility-based decision optimization, and performance-based adaptive retraining, resulting in increased decision accuracy, spectral efficiency, and labeling efficiency in wireless communication systems. Thus, the study enhances wireless systems, embedding performance-aware decision logic.