<p>Identifying and classifying lithologies is crucial for characterizing the potential of unconventional oil and gas resources. However, lithological complexity and frequent facies changes in marine–continental transitional shale systems present significant challenges in identifying and interpreting reservoir lithology. This paper focuses on the marine–continental transitional shale systems on the eastern margin of the Ordos Basin. According to the logging curves, core observations, and thin section analysis of eight cored wells, the reservoir lithology was classified into six basic types. Based on these, conventional logging curves were used as input to train machine learning models, including extreme gradient boosting (XGBoost), back propagation neural network (BPNN), support vector machine (SVM), random forest (RF), naïve Bayes (NB), and fully connected neural network (FCNN). To improve prediction accuracy, the acoustic (AC), compensated neutron (CNL), and density (DEN) logging curves that reflect changes in porosity were selected for superimposed reconstruction calculations, and new parameters (Φ₁ and Φ₂) were introduced to optimize the existing model. The study found that the accuracy, recall, precision, and F1-scores of the six classification models were all greater than 0.5, with the XGBoost model exhibiting the best overall performance, followed by the RF, FCNN, SVM, BPNN, and NB models. All original models performed well for limestone and sandstone but not for carbonaceous shale, black shale, and silty shale. With the introduction of new parameters, the revised models showed significant improvements in all evaluation metrics, with values exceeding 0.9. Among them, the XGBoost-R model performed the best overall, accurately identifying limestone and sandstone and achieving over 96% accuracy in lithology identification. In particular, it exhibited the highest identification rate for the easily confused carbonaceous shale, black shale, and silty shale. Predictions for the lithology of the Shan 2³ in the study area revealed that the central–eastern and southern regions possess excellent source rocks, indicating significant development potential. This study provides effective technical support for lithology identification of marine–continental transitional shale reservoirs and improves the efficiency of regional exploration and development.</p>