Comparison of time series forecasting methods
Mariya Shirokopetleva, Olha Ponomarenko, Zoia Dudar · Bionics of Intelligence · 2018
The article is devoted to the description and comparison of time series forecasting models and identifying the possibilities of using various models for solving forecasting problems with different initial data: time intervals, presence of seasonality and / or trends. In addition, among the two popular ARIMA and ANN forecasting methods, a more detailed review and practical comparison was made using the example of real time series of the cost of rye bread in Ukraine, forecast errors for the short and long term were identified. The estimation of the error was carried out using the software system presented in this article by comparing the deviation of the prediction results from the real data for the last period, which is 1/4 of the original data. According to the results obtained, conclusions were drawn regarding the effectiveness of both methods and the potential possibilities of their future use. It is also proposed to use the presented software system for extrapolating series in various fields, such as economics, technical systems, education, natural and social systems.