Round-Trip Time Estimation Using Regularized Extreme Learning Machine

In today's world, the availability of reliable Internet connectivity is essential. The Internet has become indispensable for activities such as e-commerce, corporate meetings and education. Many applications and services, such as video conferencing, media streaming and online gaming, depend on high-quality network performance. Round-trip time (RTT) is a metric used to assess network quality, measuring the time it takes for a packet to travel from the sender to the receiver and for the receiver to acknowledge that the data has been properly received. A high RTT indicates sub-optimal network performance and inefficient traffic, resulting in a poor user experience. Accurate RTT estimation is important for optimizing the Retransmission Timeout (RTO) value, which determines the amount of time to wait before a packet is considered lost and retransmitted. This research proposes an improved RTT estimation model by overcoming overfitting and integrating practical algorithms, using the Regularization Extreme Learning Machine (RELM) method to handle outliers, and improving efficiency by developing a regularization constant selection algorithm. This research aims to contribute to more efficient and reliable network performance management. The results of RELM's accuracy in predicting RTT are very good with an accuracy of 99.88% for validation data and 99.93% for test data. It can therefore be concluded that RELM is quite capable of predicting RTT values.