The lithology of volcanic rocks is diverse and the log characteristics are complex. Efficient lithology identification can improve the prediction efficiency of high-quality reservoir and reduce the exploration cost, thus laying a foundation for the efficient development of volcanic oil and gas resources in the later period. Aiming at the problems of low identification accuracy and complex model in the process of traditional machine learning lithology identification, this paper takes volcanic rocks in Wangfu fault Depression of Songliao Basin as the research object, comprehensively analyzes the geological characteristics of volcanic rocks reservoir, uses the corrected lithology data as the lithology sample label, and uses principal component analysis to screen out four characteristic logging curves sensitive to volcanic rock lithology identification as input. The lithology identification model is constructed by XGBoost algorithm to identify the lithology of volcanic rocks. After the lithology identification results are given by the model, the identification results are compared with those of random forest, KNN and SVM algorithms. The results show that the accuracy of XGBoost algorithm is 96.13%, while the accuracy of random forest, KNN and SVM algorithm is 93.15%, 91.68% and 91.24%, respectively. XGBoost algorithm can improve the accuracy of identification results by overfitting regularization term control algorithm, and improve the operation efficiency of the algorithm by multithreading parallel operation. The lithology identification model based on this algorithm can provide technical support for solving the problem of efficient lithology identification of volcanic rocks.