Abstract: Due to the diversity and complexity of geological conditions,the acquisition of geomechanical parameters is time-con-suming and labor-intensive and the process is very difficult. The geomechanical parameters play an important role in geotechnical anal-ysis and seismic evaluation of buildings. Based on a large number of geotechnical tests and geotechnical investigation reports in Kun-ming,this paper extracts four types of geotechnical types,and extracts four eigenvector input SVM classifiers for training,and estab-lishes the mapping relationship between soil mechanics parameters and standard penetrations. A method based on genetic algorithm to optimize SVM classifier is proposed. It is applied to the inversion between geomechanical parameters and standard penetration num-ber,and compared with the experimental results of particle swarm optimization(PSO)to optimization SVM classifier. The results show that the GA-SVM based method is very fast in geotechnical engineering parameter inversion and has strong practicability and general-ization ability.