Depressive State Detection Model in Arabic User-Generated

Abstract

One of the most well-known mental health disorders around the world is depression, affecting people's personal, professional, and social life. It is difficult for a person to be diagnosed with depression unless he goes to a psychiatrist. In our Arab society, it is difficult for a person in our Arab culture to believe in the idea of going to a psychiatrist due to the customs, traditions, and ideas of eastern Arab societies. Therefore, we found it essential for a depressed person to be diagnosed in an advanced period before he commits suicide. We found that social media (SM) is now considered one of the open societies in which the individual spends most of his day and writes about everything he feels. If the publications he records are tracked, through the text we can diagnose him as depressed or not. We used two models in this work, first we make a binary classification in which Machine Learning (ML) techniques are used, by using tweets to identify whether the tweet is expressed depression or not, ML techniques such as Gaussian Naive Bayes (Gaussian NB), Logistic Regression (LR), Support Vector Machine (SVM), Random Forest Classifier (RF), and Deep Learning (DL) use Multi-layer Perceptron classifier (MLP), LR makes the best accuracy 91%. In the second model, we used multi-classification which takes a depressing tweet from the first model and classifies it into nine classes, this was done by using DL, especially MLP networks which achieved an accuracy of 0.97.