A decision support system based on machine learning and time series for demand forecasting

Developing a real life modeling application can be quite difficult. Modeling and predicting time series can be complicated by the linear and nonlinear patterns they contain. In such cases, the use of hybrid models is prominent in order to obtain better results in terms of accuracy compared to individual models, as it combines the mechanisms that different techniques have in one place. In this study, a model-oriented web-based decision support system was developed for a real-life prediction problem that includes hybrid models as well as linear and nonlinear approaches. The performances of ARIMA (Integrated Autoregressive Moving Average), LSTM (Long Short Term Memory Networks) and hybrid model in the model base are shown on the example of bitcoin value estimation. In addition, the performances of the methods are presented comparatively for the sample datasets created based on real life application problem and the components of the developed decision support system are explained in detail.