关键词: 人机交互, 预警, 动态SVDD, 支持向量, 核方法

Abstract: After reviewing the current early warning researches，it is found that current methods in these researches are unsuitable under some circumstances，and a human-machine interactive early warning algorithm based on dynamic support vector data description（SVDD） is proposed.In the algorithm，the technique of human-machine interactive can enable the early warning system accurately identify “good” and “bad” objects.And then the technique of SVDD is analyzed，and it is found that the human-machine early warning system couldn’t work because of the over large optimization scale.To solve the problem，the algorithm of dynamic SVDD is introduced to prompt the efficiency of the system by largely decreasing the optimization scale.

Key words: human-machine interaction, early warning, dynamic SVDD, support vector, kernel methods