Least squares twin support vector machine\n(LSTSVM) is a relatively new version of support vector\nmachine (SVM) based on non-parallel twin hyperplanes.\nAlthough, LSTSVM is an extremely efficient and fast algorithm\nfor binary classification, its parameters depend on\nthe nature of the problem. Problem dependent parameters\nmake the process of tuning the algorithm with best values\nfor parameters very difficult, which affects the accuracy of\nthe algorithm. Simulated annealing (SA) is a random search\ntechnique proposed to find the global minimum of a cost\nfunction. It works by emulating the process where a metal\nslowly cooled so that its structure finally ââ?¬Å?freezesââ?¬Â. This\nfreezing point happens at a minimum energy configuration.\nThe goal of this paper is to improve the accuracy of the\nLSTSVM algorithm by hybridizing it with simulated annealing.\nOur research to date suggests that this improvement on\nthe LSTSVM is made for the first time in this paper. Experimental\nresults on several benchmark datasets demonstrate\nthat the accuracy of the proposed algorithm is very promising\nwhen compared to other classification methods in the\nliterature. In addition, computational time analysis of the\nalgorithm showed the practicality of the proposed algorithm\nwhere the computational time of the algorithm falls between\nLSTSVM and SVM
Loading....