Suicidal detection in social media means the identification of suicidal thoughts in twitter from the people. As more users increase to express their distress and suicide ideation over the internet, advanced techniques of detection have become unavoidable for early intervention. To do this, the researchers employed a hybrid deep learning model that includes hybrid model and LSTM along with NLP approaches to classify the tweets as suicidal or non-suicidal. There are a total of 10,000 tweets with both suicidal and non-suicidal content in the dataset. The process includes text input preprocessing that first tokenizes, removes stop words, and applies word embeddings followed by feeding to CNN layers for feature extraction and then LSTM layers for identifying contextual connections. This hybrid model combines the ability of CNN in recognizing patterns with the contextual understanding of LSTM, and a classification accuracy of 93.92% is obtained, which is higher than that of typical machine learning classifiers as well as independent deep learning models. Additionally, a BERT model applied during testing was able to obtain a much more impressive accuracy of 95.34%, which boosted precision, recall, and F1 score by leaps and bounds. These results suggest that CNN and LSTM architectures can identify suicide well within a mental health surveillance framework, which makes the model worthwhile in mental health monitoring on social media platforms.