Estimating Base Station Traffic and Throughput Using Machine Learning Based on Hourly Key Performance Indicator (KPI) Network Analysis

This research focuses on analyzing and predicting traffic and throughput at base stations in cellular networks using machine learning algorithms. The main research area is network performance optimization in telecommunication systems. With the increasing complexity of cellular networks and the need for resource optimization, modeling and predicting network performance has become very important. A model is developed to predict traffic and downlink throughput based on Key Performance Indicators (KPIs) captured hourly from network data. The model is trained using a comprehensive dataset that includes various KPIs. Data was collected from a cellular network site in Bandung, Indonesia, over a four-month period, providing high granularity for analysis. K-Nearest Neighbors (KNN), Random Forest, and XGBoost models were implemented to forecast network parameters. The XGBoost model demonstrated superior performance, with Mean Squared Error (MSE) and R-squared (R2) values outperforming the other models. Specifically, the XGBoost model achieved an MSE of 0.485 and an R2 of 0.976 for traffic prediction, and an MSE of 12.382 and an R2 of 0.943 for downlink throughput prediction. Hyperparameter tuning further optimized model performance. The findings underscore the effectiveness of machine learning in network optimization, contributing to the advancement of 5G technologies. These results offer a promising approach for improving resource allocation and network efficiency.