In this paper, we reviewed and analyzed various time series forecasting models using data collected from IoT mobile devices. The main attention is paid to models describing the behavior of traffic in telecommunication systems. Forecasting methods such as exponential smoothing, linear regression, autoregressive integrated moving average (ARIMA), and N-BEATS, which uses fully connected neural network layers to forecast univariate time series, are covered. The article briefly describes the features of each model, examines the process of their training, and conducts a comparative analysis of the quality of training. Based on data analysis, it was noted that for the UDP protocol, the ARIMA model has the best learning quality, for the TCP protocol - linear regression, and for the HTTPS protocol - ARIMA.
Keywords: telecommunication systems, traffic analysis, forecasting models, QoS, artificial intelligence, linear regression, ARIMA, Theta, N-BEATS