Effective Prediction on Time Series Data Using Deep Learning: An Incisive Review

Abstract

Time-series evaluation explores the way in which data differs over time and optimal prediction could determine the pathway wherein the data seems to be changing. Prediction of time-series might possess significant value for business development when one possess the access to existing information with time-constituent. Concurrently, DL (Deep Learning) algorithms are capable of offering promising solution to predict time-series due to their advantages in automatic temporal learning. With extensive usage of DL based models in varied areas like science, one might be tangled as to determine the suitable model to resolve the existing issues. Moreover, these methods might be used by existing works in various forms like hybrid. Hence, this study is instigated by an intention to afford a precise review about the recent DL based models to predict time-series. Hence, this study reviews the recent research works (2018-2023) related to the use of DL based models for effective time-series prediction in various areas. Further, a comparatively assessment is undertaken by considering the conventional DL models and varied applications of widely employed DL models for efficient prediction of time-series. Lastly, the research gaps are emphasized from conventional studies with future recommendations. This would support future researches in resolving the prevailing challenges in this area, thereby attempt to bring innovations for effective time-series prediction.