Abstract: Because of the characteristics that the linear classifier can be easily extended to nonlinear one, it becomes one of the most commonly used methods in statistical pattern recognition. According to the optimal hyperplane and Electromagnetism-like mechanism algorithm, a combinational optimization linear classification algorithm is proposed, The new algorithm extracts category features from individual samples by sample training, finds the optimal classification hyperplane, designs and realizes a combinational optimization linear classification algorithm based on the improved Electromagnetism-like mechanism algorithm. Experiments show that the algorithm has good classification results, confirm the feasibility of combinational optimization linear classification algorithm.

Key words: combinational optimization, linear classification, electromagnetism-like mechanism algorithm, category feature, sample training