Machine Learning-Based Depression Detection via Twitter Metadata

Depression, a common and major disorder is affecting a large number of individuals and has emerged as the main cause of a public health problem that needs the most attention. Serious emotional disorders, behavioral problems, and physical health issues brought on by depression can have a worse and costlier impact on a person’s health and also lead to suicide. In today’s modern era, everybody is having access to social Networks through which users share their feelings, thoughts, ideas, emotions, and struggles in life on a daily basis. Thus, the content generated by the user such as posts from social networking sites (SNSs) may be used to detect the depressive disorder in a non-intrusive manner using Natural Language Processing (NLP) and Machine Learning (ML). In this work, metadata information that Twitter offers about its users, such as the user's description can provide important details about the individuals' mental health. The dataset considered here is a benchmarked depression dataset. Here we develop different machine-learning models such as LR, RC, SVM, KNN, NB, DT and RF to detect depression using Twitter’s metadata and Sentiments. SVM offers the best performance with 71.14 % Accuracy using GLOVE embedding.