Multi-Generation Network Classification Using Signal-Centric Decision Tree Algorithms

Abstract

Mobile networks like 3G, 4G, and 5G generate massive amounts of signal data, and identifying the correct network type has become increasingly challenging as environments and signal conditions vary. Understanding network-type identification is essential for optimizing coverage, improving user experience, and supporting intelligent telecom operations. Traditionally, network classification in telecom relied on simpler models like Ridge-Classification and regression Tree, Decision Tree but these approaches required significant manual effort, struggled to adapt to changing signal conditions, and could not handle the massive volumes of signal data generated by today’s mobile networks. As signal data in modern mobile networks continues to grow in volume and complexity, automation becomes essential. This work applies machine learning to automate network-type classification, improving speed, reliability, and scalability while reducing errors and minimizing human involvement. Using a signal-metrics dataset containing features such as signal strength, signal quality, data throughput, and latency, a Multi-Layer Perceptron and a Categorical Boosting Classification and Regression Tree (CART) are applied as the proposed systems, leveraging deep learning to achieve robust performance and high accuracy in classifying network types.