A meta-analysis of the implementation of ANN back propagation methods in time series data forecasting: Case studies in Indonesia

Abstract

This research aims to systematically analyze the results of the application of Artificial Neural Network type Back Propagation (ANN-BP) methods in the prediction or forecasting of time series data, case study in Indonesia. Data is collected from the results of the ANN-BP method from indexing databases, namely Google Scholar, DOAJ, and Scopus. From search results by applying eligibility criteria including (1) keywords "prediction, forecasting, ANN Back Propagation, time-series data", (2) articles published 2011-2021, (3) the amount of data (N), accuracy rate value or correlation coefficient (R), obtained 36 qualified articles. Furthermore, the results of data analysis using JASP software obtained an average ANN-BP accuracy rate of 90% and a coefficient estimate of 0.901 at intervals of 86%-94% with random effect (RE) models. Based on the moderator variables of the year of publication is obtained the information that in the interval of 2013-2015 by 81%, in 2016-2018 by 90%, and in 2019-2021 by 94%. Finally, if the data input is higher, then the better the data pattern recognition by ANN-BP.