Quantitative analysis experiment is performed based on IR CO gas quantitative analysis model for harmful gases CO emitted by vehicle. Model is set concentration ranges of 0. 5

20 %

15 groups of samples are selected. Support vector machine（ SVM） are used to build the regression model for sample data. Related parameters are optimized based on the method of improved grid search. The experimental results show that after regression analysis by SVM

compared to the traditional method of spectral absorption

regression sample significantly closer to the calibration concentrations than the experimental values. And compared with PSO algorithm

this method obtains c = 0. 707

g = 0. 5

PSO obtains c = 55. 911

g = 0. 01

the time o

f modeling by improved grid search is reduced about 40 % of PSO algorithm. SVM used in CO concentration analysis is coincide with the test requirements and regression efficiency is high.