The Comparison Between Sequential Minimal Optimization and Multilayer Perceptron Neural Network Methods in Predicting the Commodity Prices

There are several methods that can be used in prediction system. Among them are principal component analysis and sequential minimal optimization methods. These methods have their own advantages and disadvantages. The appropriate method in the prediction system model has an effect on the accuracy of the results. The use of method in predictive system is influenced by the characteristics of the data used. Therefore, this study discussed the comparison between the characteristics of the principal component analysis and sequential minimal optimization methods in predicting commodity prices. It was aimed to find out the accuracy of the two methods in predicting commodity prices. Accuracy is indicated by the mean absolute percentage error and the direction accuracy values. The methods were tested using Weka tools to predict commodity prices. The test results used 324 data of crude oil price which showed that the MLPNN method had a better accuracy rate than the SMO method, which was 60.49% versus 60.28%. However, MLPNN has the mean absolute percentage error greater than SMO, which was 1.47% compared to 1.35%. The test results showed that the sequential minimal optimization method had a better level of accuracy than the principal component analysis method.