One of the principal issues related to earthquake resistant model is determination of liquefaction - triggering potential. In this paper, it is proposed to develop a relevance vector machine (RVM), an artificial intelligence technique that uses Bayesian interference to obtain problem solving solutions for regression and probabilistic classification which aims to improve the model performance in terms of liquefaction susceptibility of the soil. Data of Cone Penetration test exploration and post liquefaction analysis during 1999 Chi-Chi, Taiwan earthquake is considered to develop an empirical model with liquefaction as output. Out of total of two hundred and seventy five (275) CPT data, one hundred and thirty four (134) cone penetration data (CPT)till a particular depth with various engineering parameters such as Total vertical stress (σ'v), Effective vertical stress (σ'v), Cyclic Stress Ratio (CSR7.5) and Cone Resistance (qc) with their known liquefaction nature were considered from earthquake affected area. Among the collected data, 94 data sets were randomly selected for trainingto develop an RVM model for determining liquefaction of soil while rest 40 data sets were used for testing the developed model. The performance of proposed model indicates a promising approach for prediction of soil liquefaction potential.