In the modern world, depression has become a prominent element in suicide instances. Many people experience depression as a result of personal or professional difficulties, and they frequently share their feelings and opinions on social networking sites like Twitter. The material provided determines whether these feelings are favorable or negative, with negative sentiments perhaps suggesting a risk of suicide. To prevent suicide, it is essential to identify and treat depression in people. This work focuses on utilizing a Twitter depression dataset to assess depression using several machine learning models, such as CNN, Bi-directional LSTM, Uni-directional LSTM-RNN, and Vanilla RNN. The effectiveness of these models is evaluated both quantitatively and qualitatively in the study, accounting for variables such as the F1-score, accuracy, precision, recall, and AROC curve. In comparison to the other models, the results show that the Bi-directional LSTM model analyzes depression with greater accuracy, suggesting promising possibilities for early identification and intervention in people at risk of depression-related problems and suicide.