Double layer interval graph model: The universal tool for data driven market analysis and forecasting
This scientific work is dedicated to the development, improvement and application of double layer interval weighted graphs (DLIG) for non-stationary time series forecasting. This model appears to be the universal and easy-to-use tool for modeling the non-stationary time series and forecasting. We observe the double layer version of the model because it's the most representative way in the sense of main idea though you can add several layers more for different purposes. The first layer of the graph is based on empirical fluctuations of system and displays the most potential fluctuations of the system at the time of system training. The second layer of the graph as a superstructure of the first layer displays the degree of modeling error and it's connected with the first layer nodes by edges. The second layer is the way of supervised training implementation with the aim of error minimization.