Different Extrapolation Methods in Problems of Forecasting

Abstract

The task of forecasting characteristics of self-similar traffic in IoT network objects with a significant number of pulsations and the property of long-term dependence is considered, which makes it difficult to forecast in practice. Using the different extrapolation method of spline-extrapolation based on linear, cubic and B-cubic spline function and wavelet-extrapolation based on Haar-wavelet, the results of forecasting of self-similar traffic are obtained. The comparison made allowed the results of traffic forecasting based on the Haar-wavelet and the linear, cubic and B-cubic spline-function using wavelet- and spline-extrapolation. This will allow you to choose one or another extrapolationmethod to improve the accuracy of the forecast, while ensuring scalability and the ability to use it for various IoT applications to prevent network overloads.