Abstarct: Seismic exploration, as a key method for identifying hydrocarbon reservoirs, requires precise interpretation of subsurface structures and properties. The creation of geological models consistent with seismic data, although non-unique and dependent on the interpreter's perspective, is considered the initial and vital step in this process. Seismic attributes play an important role in quantitative analysis and accurate structural and stratigraphic interpretation of seismic data by revealing hidden information. The introduction of new attributes and the simultaneous use of multiple attributes (both conventional and novel) lead to the production of a massive volume of data (some of which may be redundant, superfluous, or irrelevant), which can make the interpretation of subsurface geological structures time-consuming and uncertain for human interpreters. In this context, employing automated algorithms baxsed on information fusion, classification, clustering, and machine learning methods—especially unsupervised approaches—can serve as an effective solution to reduce data volume and uncertainty while increasing the accuracy of seismic data interpretation. Usually, the use of supervised methods is met with skepticism due to the need for extensive and high-quality training data, which is difficult to obtain in seismology because of geological complexities. In this study, unsupervised clustering methods were employed as the first step of interpretation to identify reflection patterns in 2D seismic data. For this purpose, various textural and non-textural attributes were utilized, and to improve the results, new textural attributes were also introduced. Subsequently, feature selection methods were applied to reduce data volume and enhance accuracy, resulting in the selection of a subset of these attributes. Finally, the Rain Optimization Algorithm, as a new mextaheuristic optimization method, was used for optimization-baxsed clustering of the selected attributes to automatically detect reflection patterns. Its performance was compared with conventional optimization algorithms such as Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). baxsed on quantitative results, the Davies–Bouldin index indicated about a 30% improvement in clustering accuracy and reflection pattern differentiation using the Rain Optimization Algorithm, demonstrating the effectiveness of unsupervised clustering methods—particularly mextaheuristic algorithms—in more accurately identifying reflection patterns. Moreover, the Rain Algorithm was able to recognize stratigraphic structures with greater overlap with the actual geological interpretation.