Abstarct: Seismic data serve as a fundamental tool for studying subsurface structures and evaluating hydrocarbon reservoir characteristics. Buried channels are among the most significant geological features in hydrocarbon exploration, as their high porosity and permeability provide favorable conditions for hydrocarbon accumulation. However, detecting these channels in seismic data remains challenging due to subsurface complexity and data noise. Conventional approaches in this field include applying various filters to seismic data, extracting multiple seismic attributes, and employing multi-attribute analysis combined with clustering methods.In this study, unsupervised machine learning algorithms were utilized to extract key seismic attributes, including amplitude, frequency, and texture. Subsequently, data dimensionality was reduced using Principal Component Analysis (PCA), through which the dominant patterns within the dataset were identified. In the next step, three unsupervised clustering algorithms K-means, Gaussian Mixture Model (GMM), and Hierarchical clustering were applied for seismic facies classification, and the obtained results were interpreted and compared. Furthermore, to validate and determine the optimal number of clusters, Silhouette index, log-likelihood, and cophenetic correlation coefficient were employed. The proposed algorithms were implemented on two seismic datasets with distinct channel patterns.The findings revealed that integrating multi-attribute analysis with unsupervised clustering methods provides higher accuracy in distinguishing channel-related facies compared to single-attribute approaches. All three clustering algorithms performed effectively; however, the Gaussian Mixture Model (GMM) exhibited superior performance due to its higher capability in modeling complex data distributions, yielding more accurate and stable results. The applied validation indices also played a significant role in determining the optimal cluster number and improving the quantitative interpretation of the results.