Abstract: In glutenite reservoirs, the complexity of logging responses due to rock granularity poses challenges for traditional lithology identification methods. With technological advancements, combining logging data with computer technology for lithology research has become a new trend. The optimized version of the gradient boosting decision tree, the XGBoost algorithm, is widely applied in lithology identification for its efficient and accurate prediction capabilities and excellent generalization performance. This study uses the XGBoost algorithm to identify the lithology of glutenite reservoirs in the M block to improve identification accuracy. By analyzing the lithological characteristics and logging responses of the Baikouquan formation reservoirs, four logging curves (GR, AC, DEN, R_T) were selected as feature variables. A total of 468 sample data sets were divided into training and testing sets in 4： 1 ratio, and the key parameters of XGBoost were optimized through cross-validation, determining the optimal values for iteration times, learning rate, and other model parameters. The experimental results show that the XGBoost algorithm performs well in lithology identification, achieving a final accuracy rate of 91.05%, an improvement in both accuracy and efficiency compared to the C4.5 decision tree algorithm. The study results demonstrate the effectiveness of the XGBoost algorithm in improving lithology identification accuracy, providing guidance for the exploration and development of glutenite reservoirs.

Key words: XGBoost algorithm, glutenite reservoir, lithology identification, logging evaluation, Baikouquan Formation Reservoir, Mahu Depression