The rapid evolution of 5G mobile networks, ensuring stable and high-quality Internet connectivity has become critical. Radio Link Failure (RLF), caused by environmental factors such as weather conditions, wind, and surrounding physical infrastructure, is a significant challenge that can disrupt communication reliability. This research aims to predict the occurrence of RLF events using machine learning (ML) techniques to enhance 5G communication performance. By incorporating novel preprocessing and feature engineering methods, the study employs a decision tree model trained on comprehensive datasets to predict RLF occurrences not only for the immediate future but also for up to five days ahead. The prediction model integrates historical RLF data with weather forecast information from local weather stations to account for the impact of environmental changes on radio link stability. In addition to the decision tree model, various algorithms, including Random Forest, Logistic Regression, K-Nearest Neighbors (KNN), and Naive Bayes, were explored for comparative analysis of prediction accuracy. The proposed system provides a low-cost, reliable solution for improving 5G network reliability, ensuring increased capacity, and reducing latency, which are vital for the growing demands of modern communication systems.

Keywords:

5G Communication, Radio Link Failure (RLF), Machine Learning, Prediction Model, Decision Tree, Feature Engineering, Random Forest, Logistic Regression, K-Nearest Neighbors (KNN), Naive Bayes, Weather Forecast, Network Reliability, Environmental Factors, Latency Reduction, Data Preprocessing, Internet Connectivity.