Best Architecture Recommendations of ANN Backpropagation Based on Combination of Learning Rate, Momentum, and Number of Hidden Layers

This article discusses the results of research on combination learning rate values, momentum, and number neurons in hidden layer ANN Backpropagation (ANN-BP) architecture using meta-analysis. study aims to find out most recommended values at each momentum interval, namely [0.1], as well used during data training process. We conducted a meta-analysis use rate, ANN-BP. The eligibility criteria 63 include 44 complete data, 30 45 data. analysis showed that value was intervals 0.1-0.2 with RE model 0.938 (very high), 0.7-0.9 0.925 input smaller than 0.932 high). recommendation is obtained from JASP by looking effect size accuracy level sample