Hybridization of Local Search Optimization and Support Vector Machine Algorithms for Classification Problems Enhancement

Abstract

AbstractThe classification technique is defined as a task of supervised learning aims to identify the set for which a new input element is classified. SVM is a classifier, which is simple in structure, easy for training and often applied in classification problems. In the paper introduced, we improved SVM model that employs local search (LS) to enhance the accuracy of the classification. The proposed approach was experimented by eleven standard benchmark datasets from UCI repository. The results show enhancement in classification accuracy of proposed SVM model in comparison with that of the conventional SVM model.Our proposed new technique combines local search optimizer and support vector machine classifier as a reasonable solution to handle classification problems. SVM was utilized and then optimized by investing the LS optimization technique. Results of experiments drove using 10 benchmark datasets showed that the proposed LS-SVM outperformed the original SVM on all datasets. Also, when comparing the accuracy of LS-SVM to the accuracy of other approaches detailed in the literature, LS-SVM’s classification accuracy was the highest on most of eleven datasets.KeywordsMachine learningSupport vector machine (SVM)Local search (LS)ClassificationOptimization