Predicción de congestión para tráfico multimedia usando modelos de Inteligencia Artificial

Abstract

Abstract: This study evaluates the performance of predictive models based on artificial intelligence for the proactive detection of congestion in multimedia traffic networks. Simulations were developed in the NS-3 environment, configuring different levels of load and bandwidth in order to obtain quality of service (QoS) metrics such as delay, jitter, packet loss, and bandwidth utilization. From the generated data, a training dataset was built and used to validate the Random Forest, XGBoost, and LSTM algorithms. The results showed that the Random Forest model achieved the best performance, with an accuracy of 99.22% and a recall of 99.6%, confirming its effectiveness in the early prediction of congestion states. This study provides a reproducible experimental approach that makes it possible to anticipate critical network conditions, optimize resource management, and improve quality of service in multimedia traffic environments, contributing to the development of proactive management strategies based on artificial intelligence.