Depression is a mental health illness marked by profound sadness, diminished interest, and a range of emotional and physical symptoms. The illness can influence an individual’s cognition and behavior, potentially disrupting daily tasks. This is particularly pertinent in the digital age, notably on social media platforms like X, where users can candidly articulate their emotions. This study seeks to identify sadness in X users by employing the BiLSTM-GRU approach, optimized through a Genetic Algorithm, alongside FastText based feature expansion to enhance detection accuracy. The features are augmented with FastText to encompass word variants and forms, enhancing data representation. The dataset comprises 111,526 IndoNews articles and 58,115 Tweets, from which TF-IDF extraction and FastText feature augmentation were employed. The data was categorized utilizing BiLSTM, GRU, and a combination of BiLSTM and GRU. The findings indicate that BiLSTM attained an accuracy of 82.78%, reflecting a 3.04% improvement over the baseline, while GRU reached 81.93%, marking a 2.09% enhancement. The combination of BiLSTM and GRU yielded the best accuracy of 83.29%, representing a 3.48% rise. This discovery validates that a hybrid methodology including feature extraction, feature expansion, and hyperparameter optimization can significantly enhance the accuracy of depression identification. This methodology facilitates the creation of automated systems that accurately identify symptoms of depression on social media, promoting early intervention and enhancing mental health awareness within the digital community.