VKOPP: A kind of long-term prediction model for electronic system

In this paper, a novel prediction method named VKOPP for electronic system is presented, which utilizes the advantages of Volterra series and OPELM algorithm. Firstly, this algorithm in Volterra series modeling uses the minimum entropy rate method mentioned to simultaneously optimize the embedding dimension and delay time. Secondly, taking advantage of the variable selection and difference calculation to further improve the accuracy of long-term prediction. Thirdly, putting forward a novel approach based on k-Nearest Neighbor adaptive least-square method (KALE) method to replace classical LSE method to calculate the output weights of OPELM. Finally, for different kinds of activation function Taylor expansion, by comparing the polynomial coefficients, so as to calculate each order kernel functions of Volterra series. Results for both computational time and accuracy (MSE and NRMSE) are compared to other seven typical machine learning algorithms, and the experiment results are promising. In addition, this model can predict the radio frequency (RF) low-noise amplifier circuit´s stability parameter with small error.