Shear wave velocity is a fundamental parameter for geophysical, geomechanical, and petrophysical studies. To date, many wells are absent in shear wave velocity measurements due to the high cost and time-consuming to acquire. Many researchers have recently implemented a machine learning approach to estimate shear wave velocity because of its robustness in predicting a non-linear paradigm. However, many previous studies neglect the importance of optimizing machine learning's hyperparameter as many preferred to configure the hyperparameter manually, which can be less efficient, expensive to evaluate, and a time-consuming process. Optimizing the hyperparameters of machine learning is vital to obtain the maximum predictive potential. In this study, The Tree Parzen Estimator (TPE) Bayesian optimization algorithm was implemented to automatically fine-tuned the hyperparameters of Extreme Gradient Boosting (XGBoost), Random Forest (RF), and a Multi-Layered Perceptron Neural Network (MLPNN) algorithms. Subsequently, The effect of tuning hyperparameters on the performance of the technique is studied. Grid Search (GS) and Random Search (RS) algorithms are used to compare and evaluate the TPE-Bayesian optimization algorithm's performance. Common empirical relations for estimating shear wave velocity were also calculated to compare the performance between the empirical and machine learning approach. The results revealed that the TPE-Bayesian optimization algorithm managed to optimize the machine learning's hyperparameter and significantly improve the machine learning model's accuracy. Besides, the MLPNN algorithm optimized by the TPE-Bayesian optimization algorithm was able to outperform other presented methods. When computing power is limited, XGBoost with the implementation of TPE-Bayesian optimization is recommended. © 2022, Petroleum and Coal. All Rights Reserved.