The primary objective of geophysical research is the exploration of underground structures and to serve as a valuable tool for geological interpretation. The formation structure and properties can be determined by analyzing the physical properties of the underground medium reflected by geophysical data, such as density, velocity, magnetic susceptibility, resistivity, and more. Given the numerous solutions of a single geophysical method, comprehensive geophysical interpretation is currently a feasible and effective approach. This study explores lithology prediction, providing a summary of the basic principles and steps of comprehensive geophysical interpretation methods for lithology prediction. Additionally, it outlines the main technical methods of comprehensive lithology prediction involving two kinds of technical routes: knowledge-driven and data-driven. The knowledge-driven method uses prior information. It is simple, direct, and easy to understand, but has weak adaptability to the complexity and high dimension data. The data-driven method employs a mathematical statistics strategy to explore the relationship between data and has a robust capacity to adapt to complex scenarios. In solving practical problems, the supervised machine learning method, based on sufficient rock physical properties research, not only incorporates prior knowledge but also maximizes its internal data exploration ability. It can enhance the accuracy of lithology prediction and interpretation, better establish the corresponding relationship between geophysical and geological information, and support the exploration needs of resources and energy.