Artificial neural networks, especially thebackpropagation method are
widely used to solve various problems. In artificial neural networks, the important
thing that determines its performance is the training algorithm used.The
performance of the training algorithm is said to be optimal in providing solutions
can be seen from the errors generated by the network. The smaller the error
generated, the more optimal the performance of the algorithm. In the previous
study, the most optimal training algorithm based on the results of the smallest
error using 5 neuron inputs, 10 neurons in 1 hidden layer, 1 neuron output with
the test level α = 5% is the Levenberg-Marquardt algorithm with an average
error rate of 0,0002196. In this research, 12 training algorithms were tested to
find out the most optimal algorithm was taken from the smallest error rate. This
study uses a mixed method, namely development research with qualitative and
qualitative (using statistical tests). The research data sources used were random
data with 10 neuron inputs, 14 neurons in 1 hidden layer, 1 neuron output with a
learning level of 0.01, 0.05, 0.1, 0.2, 0.3, 0.4 , 0.5, 0.6, 0.7, 0.8, 0.9, 1. The
conclusion of the study is that training algorithms in backpropagation networks
that have the smallest error (most optimal) with network parameter control target
error = 0.001, maximum epoch = 10000, learning rate (lr) = 0.2 is the
Levenberg-Marquardt algorithm with average rates error of 0,00010132106600.