Shear wave velocity and compressive wave velocity provide valuable information about mechanical properties of a formation. Because of high costs of collecting shear sonic logs, such information is only collected for few of wells in an oil field. Several attempts have been made to estimate shear wave velocity based on petrophysical logs. Because several factors influence shear waves, proper estimate of compressive wave on even a simple model needs examination of input parameters. Literature review showed that there is a paucity of similar works in this regard. Using bivariate regression, the relationship between each one of the input logs and the target log was examined. Afterward, stepwise multiple regression, multi-purpose optimization algorithm (NSGA-II), and MLP neural network were used to examine input parameters and to choose the proper input parameter. To avoid optimum local answer by the neural network, particle swarm optimization was used to optimize the parameters. The results indicated that regardless of the method, increase of the inputs decreased estimated error and error reduction rate followed a descending trend with an increase in the number of inputs. The error reduction rate by artificial intelligence was more than that of regression models. Error reduction rate of the model with one and two input variables was higher than other modes in the both methods. The results also confirmed higher performance of NSGA-II algorithm in choosing the input parameters. In the case of several input parameters, the algorithm can be used to determine the input parameters.