Comparative analysis of prediction methods of stationary and nonstationary series

The basic methods of analysis and forecasting of time series are considered. The effectiveness of the method of autoregression, the model of autoregressive integrated moving average (ARIMA), the method of singular spectrum analysis (SSA) and methods using neural networks based on deep learning for forecasting of time series was investigated. The effectiveness was studied on the example of the values of the function of the sinusoid, the minimum daily temperature, the market index S&P 500 and the daily number of passengers of the airline on the criterion of the lowest mean square error. A comparative analysis of the obtained results is carried out and the feasibility of using these methods is substantiated.