Particle Swarm Optimization for Support Vector Machine Model Parameter Tuning

This file provides MATLAB code that implements Particle Swarm Optimization (PSO) to optimize parameters for a cost-sensitive Support Vector Machine model, specifically designed for handling imbalanced datasets. The implementation focuses on optimizing three critical parameters: C1, C2, and sigma, which control the trade-off between margin maximization and classification error. Support Vector Machine is a widely-used supervised learning algorithm in machine learning that demonstrates excellent performance for imbalanced data classification. However, SVM performance is highly dependent on proper parameter selection. Our implementation uses Particle Swarm Optimization to automatically find optimal parameter configurations through iterative swarm intelligence. The algorithm maintains a population of candidate solutions (particles) that navigate the parameter space based on their individual best positions and the global best position found by the swarm. This MATLAB code provides a complete implementation framework including: particle initialization with random positions and velocities, fitness evaluation using SVM classification accuracy, velocity update equations incorporating cognitive and social components, and position updates with boundary handling. The code features parameterizable swarm size, iteration count, and inertia weight settings. Users can easily adapt this implementation to optimize their SVM models by modifying the dataset loading section and adjusting the parameter search ranges, achieving improved classification performance through automated parameter tuning.