Time Series Forecasting by the Arima Method

Abstract

The variety of communication services and the growing number of different sensors with the appearance of IoT (Internet of Things) technology generate significantly different types of network traffic. This implies that the structure of network traffic will be heterogeneous, which requires deep analysis to find the internal features underlying the data. A common model for analyzing the processes of a multiservice network is a model based on time series.
 Numerous empirical data studies indicate that the packet intensity time series do not belong to the general aggregates of a normal distribution.
 The problem of predicting network traffic is still relevant due to managing information that flows into a heterogeneous network. 
 In this work, the authors studied the time series for stationarity in order to select an appropriate forecasting model. A visual assessment of the series assumed non-stationarity. The Augmented Dickey-Fuller Test is applied, and the measured network traffic is predicted using the ARIMA (Auto-Regressive Integrated Moving Average) statistical method. Results were obtained using the Econometric Modeler Matlab (R2021b) application. The results of the autocorrelation function (ACF) and partial ACF are analyzed, with the help of which the ARIMA model is optimized. As a result of the study, a software algorithm for the ARIMA (0,2,1) model was developed.

Optimization of Demand Forecasting in the Supply Chain Management of Apparel Industry

“…Time series forecasting using ARIMA is a fundamental method, which is widely applied in many elds of science, including short term and longterm predictions. ARIMA's adaptability to different forecasting periods reveals its effectiveness in precise forecasts (Bektemyssova et al 2022).…”

Fuzzy Model for Time Series Forecasting

“…Predictive data provide the necessary information to solve the problem of managing information flows in the network. Time series modeling is one of the ways to predict them [2].…”