A novel. to the best of our knowledge, technique based on hierarchical agglomerative clustering (HAC) is proposed to classify the direct and obstructed links in intersatellite optical wireless communication (IsOWC) systems. Prior to the training phase, an exploratory data analysis and preprocessing technique are conducted on a dataset composing 250 instances with four unlabeled input features: relative intensity noise, propagation distance, pointing error, and insertion loss. These datasets extract from IsOWC systems utilizing on-off keying modulation. The HAC technique demonstrates exceptional performance, achieving a classification accuracy of 92.7%, surpassing other machine learning techniques. Additionally, the impact of input feature combinations is discussed in detail using dendrogram plots and various performance metrics. The results provide valuable insights for localizing satellite constellations in earth orbit and advancing global Internet accessibility.