финансовые временные ряды

Introduction. Data Mining algorithms enhancement leads to the solution of predictive analytics tasks in more efficient ways. The models ensembles are one of the actively developing areas, especially in those branches where predictive accuracy is more important than the interpretability of the model. Theoretical analysis. The forecasting problem requires careful study of the data set and methods suitable for analysis. It includes the solution of such subtasks as the choice of a forecasting model, the analysis of the forecast accuracy. Models ensembles are used to combine the predictions of several basic models in order to reduce forecasting errors and increase the generalizing ability of the individual models. Conceptual schemes are presented for the basic ensemble methods (bagging, boosting, stacking, blending). Empirical analysis. The practical aspects of forecasting financial time series, implementation of several models ensemble for forecasting problems are explored in the article. Results. To build models ensembles, we developed an application with a web interface that provides the ability to evaluate models with different error metrics, choose a more accurate models ensemble. With the web application, users can select and configure models to form forecasting ensembles, test the resulting models, and visualize the results. Testing of the models ensemble for the analysis of share prices time series is presented.