The main role of drilling optimization is a decrease in the drilling cost and non-productive time (NPT) for drilling operations. The penetration rate directly influences the overall cost and cost per foot of drilling operation. Thus, the penetration rate prediction and optimization for drilling wells is one of the most crucial parameters to enhance drilling efficiency. Normally, physics-based ROP modeling is widely used to predict bit response or investigate ROP by using nearby offset data. Due to the complexity and nonlinear of ROP, and the confidence level of ROP models with low R squares, data-driven modeling such as machine learning (ML) has become a more attractive study. This paper has been developed on ROP models using artificial neural network (ANN) and compares the results of physics-based ROP models such as the Maurer model, Bingham model, Warren model for perfect cleaning model, Warren model for imperfect cleaning model, and multiple regression based on the significant level of correlation coefficients of R square from models. Drilling Oligocene formations on 8-1/2’’ hole sections have been collected from six drilled wells in the continental shelf of offshore Vietnam. The ROP prediction results were obtained from the ANN model compared with physics-based models. This comparison has shown that the predictive ROP of the power ANN model with an R square confidence level is higher than that of physics-based models.