In order to effectively utilize logging data to improve the gas layer identification accuracy of low-porosity, low-permeability sandstone reservoirs, this paper focuses on the Shanxi Formation reservoirs in the eastern Ordos Basin as the research object, and, based on a review of the literature, conducts detailed studies on gas layer identification in combination with actual logging data.Regarding the reservoir characteristics of the study area, conventional identification methods exhibit poor accuracy and a high misjudgment rate in recognizing gas layers. This paper introduces a Northern Goshawk Optimization-based Random Forest model (NGO-RF) to conduct a refined identification of gas layers.First, the core parameters of the random forest model (number of decision trees, minimum number of leaves) are optimized using the NGO algorithm. Then, the optimized parameters are applied to the random forest model to complete the prediction process for atmospheric layer identification.To verify the effectiveness of this model, the study selected three commonly used models for comparison: the unoptimized random forest model, the particle swarm optimized support vector machine model, and the BP neural network model.The research results demonstrate that the NGO-RF model exhibits the best performance in gas layer identification in the study area, achieving an accuracy of 99.45% on the training set and 96.25% on the prediction set. Both accuracies surpass those of the other three comparison models, fully confirming the suitability of the NGO-RF model for gas layer identification in low-porosity, low-permeability sandstone reservoirs of the Shanxi Formation in the eastern Ordos Basin.