FORWARD AND BACKWARD FORECASTING ENSEMBLES FOR THE ESTIMATION OF TIME SERIES MISSING DATA

Abstract

The presence of missing information in time arrangement is large hindrance to the fruitful execution of estimating models, as it prompts a critical decrease of helpful information. In this work we propose a multipleimputation-type structure for assessing the missing estimations of a period arrangement. This structure depends on iterative and progressive forward and in reverse guaging of the missing qualities, and building outfits of these gauges. The iterative idea of the calculation permits reformist improvement of the estimate exactness. What's more, the distinctive forward and in reverse elements of the time arrangement give helpful variety to the outfit. The created system is general, and can utilize any hidden AI or customary determining model. We have tried the proposed approach on huge informational collections utilizing straight, just as nonlinear hidden estimating models, and show its prosperity.