Text categorization using SVM with exponent weighted ACO

Support Vector Machine is a powerful tool for non-linear high-dimensional classification problem such as text categorization. Parameters include balance parameter C and kernel function parameter σ play important roles in Support Vector Machine. However, classic method which selects parameters manually will restrict the improvement of classifying performance when using Support Vector Machine. This article proposes an exponent weighted algorithm to overcome local optimization and low convergence rate problems in Ant Colony Optimization. The novel Ant Colony Optimization algorithm is implemented and be used to optimizing parameters of Support Vector Machine in a Chinese text categorization system. The experimental results reveal this method has a higher precision and efficiency than traditional Support Vector Machine based Text categorization systems.