In the last decade and so, the use of social media is becoming a part of daily life, especially for the young generation; its’ use is at their fingertips. This extensive use of Social Network Services (SNS) is generating a huge volume of data by sharing thoughts in form of text, images, audio, and videos. The activities done by individuals at these SNS suggest many things about their personalities. Medical health scientists are interested to analyze these activities to find out the state of mind and depression disorders in human beings. Sometimes this state of mental depression disorder (MDD) can lead to a big loss. A timely evaluation can suggest initiating preventive measures to avoid devastating consequences. To support such people who are suffering from depression disorder more congenial methods are needed which can analyze and recognize such depressive activities. In this work, various sentiment-based algorithms are applied to detect depression signals from social media posts. A comparative study is presented to evaluate the efficacy of various standard machine learning algorithms in detecting depression signals. We also applied some regression-based techniques to study the barrier losses in machine learning algorithms. It is found that the Support Vector Machine (SVM) Classifier outperformed the other counterparts and produced an average accuracy of 92 percent for classifying suicidal (depressed) tweets from non-suicidal (Nondepressed) tweets.

Depression Detection, Emotion Analysis, Sentiment Analysis, Opinion Mining, Social Network Analysis, Microblogging, Twitter, Mental Illness Detection, Natural Language Processing.