Least Squares Support Vector Machine Regression Based on Sparse Samples and Mixture Kernel Learning

Least squares support vector machine (LSSVM) is a learning algorithm based on statistical theory. Itsadvantages include robustness and calculation simplicity, it has good performance in the data processingof small samples. The LSSVM model lacks sparsity unable to handle large-scale problem, this articleproposes an method mixture kernel sparse This reduces theinitial training set sub-dataset using selection strategy. It converts single function theLSSVM into mixed optimizes its parameters. reduced used fortraining LSSVM. Finally, group of datasets UCI Machine Learning Repository were verify theeffectiveness proposed algorithm, which applied real-world power load achieve better fittingand improve prediction accuracy.