SECOS is one of the artificial neural networks that can be used in this study to make a prediction or forecasting of gold prices. This research will produce a combination of parameters, namely learning rate 1, learning rate 2, error threshold that will be carried out on the system then in the training process the data is quite influential on the resulting error value. From the results of the combination of parameters and testing with the SeCos algorithm shown in Figure 4.2, the smallest error value at layer 3 is 54262,375, which is obtained from the learning rate 1 parameter is 0.5, learning rate 2 is 1, learning rate 3 is 1.5. while the largest error value is 46023.9375. The results of the SECOS algorithm in forecasting the gold price can run well, which shows that the model from the implemented training and testing data can predict gold prices.