Training algorithm is the most important part in Artificial Neural Network (ANN). The performance of this algorithm is influenced by several network parameters including the number of neurons in the input layer, the number of neurons in the hidden layer, maximum epoch, learning rate (lr). In a previous study testing use a neuron model 6-10-1 the test results obtained that the Levenberg-Marquardt (trainlm) algorithm is the most appropriate algorithm in recognizing data pattern with an average level of appropriateness of 87.5%. In this study, testing the level of accuraccy of data pattern recognition training algorithm on artificial neural network use the neuron model 10-16-1 and 15-29-1, the networks parametes used between error targets = 0.001 (10-3), maximum epoch = 10000 (104), the value of learning rate (lr) = 0.01, 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0. The result of the research carried out in neuron model 10-16-1ANN training algorithm whichs is the most optimal level accuracy of data pattern recognition in terms of the smallest delta average and the percentage accuracy of data pattern recognition is Levenberg-Marquardt (trainlm) training with the value the average delta = 0.00632000000 at the learning rate (lr) = 0.9 whit a percentage of compatibility recognizing the data pattern of 100%. In the model 15-29-1 ANN training algorithm the most optimal level of accuracy of data pattern recognition in terms of the smallest delta average and the percentage match for the average training algorithm delta at each learning rate (lr) is Levenberg-Marquardt (trainlm) with the average value of data = 0.00530500000 at the learning rate (lr) = 0.7 with the percentage of accuracy in recognition the data pattern of 100%.