Due to the drastic drop in the oil price, the priority goals of oil operating companies become increasing the drilling efficiency and reducing the overall drilling cost. The ultimate real-time drilling efficiency can be achieved by optimizing the drilling processes. One of the drilling processes that can lead to remarkable improvement in drilling efficiency is drilling hydraulic. Prediction of hydraulic pressure for drilling operations is essential, e.g., for well planning before and decision support while drilling. By considering the whole picture, drilling hydraulics can determine whether or not the well will reach the planned depth. Moreover, non-productive-time (NPT) reduction in drilling is still a significant concern as drilling a well is often the most expensive process. Planning and design of drilling hydraulics help reduce non-productive time and cost. Therefore, it is crucial to create a well-rounded hydraulic design. The optimization goal is to maximize the pump's power to help the bit drill at maximum efficiency by minimizing the energy loss due to friction in the circulating system and using the saved energy to improve bit hydraulics. Over the past decade, several methods and techniques have been proposed aiming at optimizing drilling hydraulic in real-time; one of these techniques is machine learning, which has shown promising results in terms of prediction and optimization. Nevertheless, the real-time implementation of these techniques is still challenging since most of the published work tried to perform prediction tasks rather than the optimization task. In this regard, this thesis tries to tackle the shortcomings of the recently published related methods by proposing a holistic approach for real-time drilling hydraulic optimization. This thesis is divided into two main parts. In the first part, the advantages and limitations of current approaches used for drilling hydraulic optimization, including modeling-driven approach, numerical simulation, and data-driven approach application are pointed out. The second part is dedicated to explaining in detail the methodology followed to develop the holistic approach. The outcome of the master thesis is a standalone application based on two modules. The first module implies a data-driven predictive model with two implemented machine learning (ML) algorithms that provide a real-time prediction of standpipe pressure (SPP) and annular pressure losses (APL). The second module consists of an optimization algorithm that uses the predicted values of SPP, APL, and calculated drill string pressure loss to generate the combinations of optimized surface drilling parameters that lead to the highest hydraulic efficiency based on preselected optimization criterion. To evaluate and determine the shortcomings of the developed holistic approach, case studies were carried out. The final results of the case studies reveal that the machine learning techniques can be used to solve the real-time hydraulic optimization problem.