CeDA-BatOp 2.0: Enhanced Framework for Base Station Parameter Optimization and Automation with Joint Optimization, Controlled Drift Analysis and Pseudo-Labeling

In the ever-evolving landscape of cellular networks, the pursuit of optimal network performance remains a constant endeavor. This paper introduces CeDA-BatOp 2.0, an enhanced framework designed to address the complexities of network optimization. Building upon the foundations of its predecessor CeDA-BatOp version 1.0, this upgraded version presents a Multi-task learning (MTL) approach that leverages joint training of a base station parameter predictor and clustering. By intertwining these tasks, CeDA-BatOp 2.0 demonstrates its ability to provide a holistic and efficient solution for network optimization. Furthermore, this framework introduces an innovative analysis of controlled data drift using Gaussian Mixture Models. The controlled drift analysis allows the framework to adapt to real-world data variations effectively, enhancing network performance and reliability. It also incorporates a pseudo-labeling strategy using Multi-View Co-Training, streamlining the retraining process. This feature is crucial in scenarios where manual labeling is impractical, making the system adaptive and efficient in responding to changing network conditions. Through the simulated data, the paper demonstrates the framework’s efficacy, achieving significant improvements in network optimization. CeDA-BatOp 2.0 is a step forward in addressing the complexities of cellular network optimization. By combining MTL, drift analysis, and pseudo-labeling, it provides a comprehensive solution that adapts to the dynamic nature of cellular networks, ultimately enhancing network performance and the user experience.