TY - GEN
T1 - Accuracy rate of ANN back propagation architecture with modified algorithm
T2 - 3rd International Seminar on Science and Technology: Science, Technology and Data Analysis for Sustainable Future, ISSTEC 2021
AU - Syaharuddin,
AU - Fatmawati,
AU - Suprajitno, Herry
N1 - Publisher Copyright:
© 2023 Author(s).
PY - 2023/5/25
Y1 - 2023/5/25
N2 - This study aims to analyze the accuracy of artificial neural network back-propagation (ANN-BP) architectures in forecasting time series data using modified and unmodified algorithms. The ANN-BP method is suitable for data with a prediction matrix size of m × n to facilitate the input of each layer. The data was sifted from articles indexed by databases such as Scopus, Sciencedirect, DOAJ, and Google Scholar for the 2010-2021 publication year. The Meta-Analysis method was used to determine the accuracy rate for each case. Based on the results of data analysis using JASP software that met the inclusion and exclusion criteria, the information obtained showed that the average accuracy of ANN-BP was 85% based on random effects (RE) models including test results based on moderator variables; for the modified algorithm, the accuracy is in the interval of 0.77-0.93, and if the algorithm is modified, the accuracy is in the interval of 0.81-0.90. These intervals suggest that modifications to the ANN-BP algorithm can improve accuracy rates. These results were obtained from a meta-analysis of 26 articles data that met the criteria for inclusion and exclusion that have been established.
AB - This study aims to analyze the accuracy of artificial neural network back-propagation (ANN-BP) architectures in forecasting time series data using modified and unmodified algorithms. The ANN-BP method is suitable for data with a prediction matrix size of m × n to facilitate the input of each layer. The data was sifted from articles indexed by databases such as Scopus, Sciencedirect, DOAJ, and Google Scholar for the 2010-2021 publication year. The Meta-Analysis method was used to determine the accuracy rate for each case. Based on the results of data analysis using JASP software that met the inclusion and exclusion criteria, the information obtained showed that the average accuracy of ANN-BP was 85% based on random effects (RE) models including test results based on moderator variables; for the modified algorithm, the accuracy is in the interval of 0.77-0.93, and if the algorithm is modified, the accuracy is in the interval of 0.81-0.90. These intervals suggest that modifications to the ANN-BP algorithm can improve accuracy rates. These results were obtained from a meta-analysis of 26 articles data that met the criteria for inclusion and exclusion that have been established.
UR - http://www.scopus.com/inward/record.url?scp=85161436849&partnerID=8YFLogxK
U2 - 10.1063/5.0137185
DO - 10.1063/5.0137185
M3 - Conference contribution
AN - SCOPUS:85161436849
T3 - AIP Conference Proceedings
BT - 3rd International Seminar on Science and Technology, ISSTEC 2021
A2 - Sahroni, Imam
A2 - Musawwa, Muhammad Miqdam
A2 - Fauzan, Achmad
A2 - Werdyani, Sista
A2 - Zain, Zainiharyati Mohd
A2 - Fatimah, Is
A2 - Riyanto, null
PB - American Institute of Physics Inc.
Y2 - 30 November 2021
ER -