A precise model of shear wave velocity (Vs) is an important task for assessment of rock stiffness, elastic modulus, design of underground mine opening, micro seismic hazard assessment and wellbore stability analysis in hydrocarbon exploration works. The major objectives of this paper are to evaluate the effectiveness of ensemble machine learning techniques for predicting Vs, perform a sensitivity analysis on the predicting variables, and to develop an improved model to estimate Vs for sandstone rocks. A list of statistical model performance indicators such as determination of coefficient (R2), average absolute relative error, maximum relative error, and root mean square error are applied to assess the efficacy of data-driven predictive models with random forest, extreme gradient boosting and support vector regression techniques. According to these results, the most important contributing predictor variable is compressional wave velocity. An improved correlation is developed for predicting accurate Vs which is compared and verified with gas field well log data of clastic sedimentary rocks. It achieved excellent accuracy with the least statistical error and a high R2 of 99%. The proposed model strategies and correlation can be adopted for further evaluation of mining excavation design, rock strength prediction and hydrocarbon exploration activities with environmentally friendly impacts and nominal costs.