Study on Non-iterative LS-SVM Based on Entropy and Its Applications

By comparing and analysing the algorithm of non-iterative LS-SVM based on Renyi-entropy, traditional LS-SVM and SVM with second-order cone programming, this paper concludes whether the number of training samples or computing time, non-iterative LS-SVM algorithm based on Renyi-entropy are significantly better than the algorithm of traditional LS-SVM and quadratic programming and it also proves the superiority and effectiveness of applying the concept of Renyi-entropy on financial distress prediction. At the same time, by the comparison of different point of 3 years of ST which is from 1 to 2, the author concludes the forecast accuracy of 1 year ago before ST, the further distance away from the point of ST, the lower the prediction accuracy is.