Abstract—Data mining is the process to find the knowledge from the huge amount of stored information and use the discovered knowledge to predict or classify the new data item that its class label is unknown. Among many available algorithms to do data classification, support vector machine is one of the most accurate mining methods. Support vector machine is a parametric approach such that proper setting of parameter value can directly influence the classifying performance of the machine. Currently, genetic algorithm can find the best parameter for support vector machine. The genetic algorithm is the search algorithm for optimal answer with adaptive heuristic search based on the evolutionary characteristic of nature. But the problem of genetic algorithm is that sometime the algorithm cannot find the best parameter because the improper setting of a random initial value. In this research, we propose the new technique to improve performance of genetic algorithm to find the best parameter with restarting concept. We show the performance of the proposed technique with application for image-based forest type classification over the forest area in Japan with the satellite image data from the ASTER satellite. The results show that the proposed technique can classify the forest type more accurate than other existing techniques.
Index Terms—Data classification, restarting genetic algorithm, support vector machine, forest type classification.
The authors are with the School of Computer Engineering, Suranaree University of Technology (SUT), 111 University Avenue, Muang, Nakhon Ratchasima 30000, Thailand (corresponding author: K. Suksut, Ph: +66879619062; e-mail: firstname.lastname@example.org, email@example.com, firstname.lastname@example.org, email@example.com).
Cite: Keerachart Suksut, Nuntawut Kaoungku, Nittaya Kerdprasop, and Kittisak Kerdprasop, "Parameter Optimization with Restarting Genetic Algorithm for the Forest Type Classification," International Journal of Machine Learning and Computing vol. 7, no. 6, pp. 213-217, 2017.