Relative Perusal of ML Classifiers for Depression Detection in Twitter Feeds

Abstract

Depression is viewed as a significant cause of suicidal inclination. It affects the style of writing manifested in the text. Analyzing linguistic markers in social media posts help in recognizing and classifying whether thoughts or sentiments expressed in source text correspond to a depressed user. A large amount of emotion-rich data generated by social media platforms is in the form of tweets, feeds, blog posts, etc. Analysis of this user- generated data helps in understanding an individual’s state of mind. The main focus is to scrutinize the posts of users of twitter to analyze the depression attitudes of users. Natural Language Processing and ML techniques like MNB, TF-IDF, SVC, SGD, and LR have been utilized for training the data set and estimating the efficacy of our proffered approach. Firstly, the words are reduced into their morphological form during pre-processing. Then, a depression analysis model is built based on the suggested method and various features of depressed users derived from psychological research. Tweets with the hashtags #depression are classified based on their content and their relative tendencies towards depression. Tweets related to social distance, workplace stress, peer pressure, family problems, personal weakness, failure, mocking, and self-stigma helped in depression detection. The results have been rendered using the key evaluation measures, which include accuracy, precision, and F1-score. The results of the study may be beneficial in assisting mental health awareness and supporting organizations to provide data about resources and counter common depression.