Addressing suicide prevention is a critical issue on a global scale, given the notable occurrence of annual deaths related to suicide and the numerous attempts made worldwide. The rise of social media as a platform for emotional expression underscores the need for early detection of suicidal thoughts. This study introduces an innovative approach utilizing Natural Language Processing (NLP) techniques to analyze behavioral patterns and linguistic cues in social media interactions. Hence, this study introduces a pioneering ensemble optimized technique that combines the Bi-GRU and Attention frameworks. This integration significantly improves the scalability and efficiency of detecting suicidal ideation. Given the increasing complexity and the growing significance of social media platforms, employing techniques like ensemble hyperparameter tunned advanced methods are becoming progressively essential to ensure detection of suicidal posts. The proposed approach has been evaluated using two different datasets, Reddit and Twitter. The model produces the noteworthy result as accuracy of 92% and 91% for the Reddit and Twitter datasets, correspondingly. These findings underscore the approach's remarkable accuracy in classifying suicidal ideation posts, bearing significant implications for social media platforms.