Lithology identification is recognized as a fundamental basis for deep earth resource exploration and evaluation. Conventional well logging and mud logging methods are confronted with low resolution and strong ambiguity in complex geological formations, while machine learning, equipped with powerful nonlinear mapping and automatic feature learning capabilities, has been established as a mainstream technical approach for intelligent lithology identification. A systematic review of research progress in machine learning based lithology identification was conducted using well logging and mud logging data. At the characteristic indicator level, the selection rules of sensitive well logging curves and characteristic elements were summarized for sedimentary, magmatic, and metamorphic rocks, and feature engineering techniques including data cleaning, dimensionality reduction, and multi source fusion were reviewed. At the algorithm level, the application status of mainstream algorithms, including artificial neural networks, support vector machines, ensemble learning, and cluster and discriminant analysis, was evaluated, and the implementation effects of intelligent optimization strategies such as particle swarm optimization and Bayesian optimization were analyzed. A comprehensive comparison of representative algorithms was performed from the perspectives of identification accuracy, computational efficiency, and small sample adaptability. At the application level, the application status of machine learning based lithology identification in basic geological research and resource exploration is systematically reviewed. Regarding existing problems, key issues such as insufficient model generalization, limited algorithm adaptability, and weak research in deep coal measure strata are identified and discussed. On this basis, future research directions are proposed, including joint identification of stratigraphic and lithologic information, construction of cross work area general large models, algorithm improvement tailored to data characteristics, and expansion into deep coal measure resource exploration. This review is expected to provide a theoretical reference for intelligent lithology identification under complex geological conditions.