Exponential smoothing-based forecasting of self-similar internet of things traffic

Abstract

The rapid growth of internet of things (IoT) devices generate highly variable and self-similar traffic patterns, creating challenges for maintaining quality of service (QoS) in modern telecommunication networks. Accurate short-term forecasting of such traffic is essential for efficient resource allocation, yet its fractal characteristics and long-range dependence complicate prediction. This study investigates the use of simple exponential smoothing for short-term forecasting of self-similar IoT traffic by evaluating three smoothing coefficients (a=0.1, 0.5, and 0.9). The Hurst exponent (H=0.5) confirms the presence of self-similarity in the observed traffic. Experimental results show that a=0.1 provides the highest prediction accuracy, achieving a mean absolute percentage error (MAPE) of 25.82% when forecasting traffic values within a 32-minute horizon. The method effectively captures underlying trends while reducing noise sensitivity. These findings demonstrate that exponential smoothing offers a lightweight, interpretable, and practical solution for real-time IoT traffic forecasting, supporting dynamic network load management under highly variable traffic conditions.