<![CDATA[Coal price prediction is needed as support for coal user industrial to buy coal. Prediction result can be used to make next budgeting. This research uses Support Vector Regression (SVR) method to predict coal price. SVR is applied through data normalization, hessian matrix calculation, α searching through sequential learning, and regression function calculation. Kernel for hessian matrix stage can determine accuracy of prediction, so in this research Gaussian RBF kernel and ANOVA kernel are used and analyzed the effects. To obtain predictive results with good accuracy, testing of each parameter is performed and evaluated by mean absolute percentage error (MAPE). The average of MAPE for testing are 9,64% with Gaussian RBF kernel and 8,38% with ANOVA kernel, which are categorized good, on 48 training data for 12 testing data and optimal parameters are ε 0,00001; cLR 0.01; C 0.5; λ 0.5 with Gaussian RBF kernel and 1 with ANOVA kernel. SVR gives the most optimal result when predicting the next month price. The predicted results of the two kernels are not too different, but the ANOVA kernel works better on this coal price data.]]>
Copyrights © 2018
