An Efficient Model for Data Classification Based on SVM Grid Parameter Optimization and PSO Feature Weight Selection

The support vector machine (SVM) is a classifier with different applications due to its perfect experimental performance compared to other machine learning algorithms. It has been used mostly in pattern recognition, fault diagnosis, and text categorization. The performance of SVM is extremely dependent on the sufficient setting of its parameters such as SVM max-iteration and SVM kernel-type. Therefore, the choice of suitable initial parameters of SVM will result in a good performance and classification result. This paper introduces a new schema for optimizing SVM parameters using grid search and particle swarm optimization PSO feature weighting. The experimental results demonstrate that the new method obtained a high accuracy compared to the traditional SVM and other SVM-optimization methods