Abstract: To overcome the challenges of strong heterogeneity and ambiguous log responses in identifying barriers in the tight sandstone reservoirs of Oilfield LN in the Tarim Basin, this study proposes a novel method (RF-AdaBoost) that integrates the random forest (RF) algorithm and adaptive boosting (AdaBoost) ensemble learning. The method constructs a sensitive multi-feature sample set from core thin-section and log data and applies ADASYN for sample augmentation. A grid search-optimized RF model, with the enhanced hyperparameters including the depth of decision trees, the number of estimators, and feature sampling rate, is then embedded within the AdaBoost framework, where iterative weighting is employed to significantly enhance the robustness and accuracy in identifying fine-scale barriers. The results of a comparative analysis of four models: RF-AdaBoost, k-Nearest Neighbors (KNN), RF, and AdaBoost, based on log data from 7 wells in the study area, indicate that the RF-AdaBoost model achieves an average F1-score of 94.33% on the test set and outperforms KNN, RF, and AdaBoost by 9.56%, 3.13%, and 2.45%, respectively. It demonstrates strong anti-noise capability, with a performance degradation of less than 3% under 20% simulated noise. It also exhibits high accuracy in distinguishing the boundaries of argillaceous, calcareous, and petrophysical property-controlled barrier types. The identification results, integrated with cross-well sections and the injection-production relationship, reveal flow regulation effects, which are consistent with geologic features and production performance. This demonstrates the method's capability to accurately identify barriers and characterize their dual functions of inter-layer sealing and lateral fluid diversion during hydrocarbon migration. The proposed RF-AdaBoost model provides a robust solution for intelligent barrier recognition in tight sandstone formations, especially under non-coring and low signal-to-noise ratio conditions. It offers technical support for fine-scale reservoir characterization and optimized waterflood development planning.