Abstract: This paper applies the support vector machine technology to train and classify gas feature vectors and measured samples from the surface acoustic wave (SAW) sensor. The objective is to identify toxic gas from gas samples. Experimental results verify the feasibility and effectiveness of this solution.