Prediction of anxiety and depression among the transgender in Bangladesh

Abstract

Today, mental health is as important as safety. It affects people directly and indirectly. Transgender people with mental health issues are often overlooked and in our nation, especially transgender women, suffer from anxiety, depression, and suicide. This research examines depression and anxiety prediction. Classification models and NLP are used to detect depression and anxiety in 41 Bangladeshi transgender people. The data came from sociodemographic, victimization, social support, interpersonal functioning, depression, anxiety, and self-esteem questionnaires (RSE). We predicted depression and anxiety in transgender people using culturally adapted depression and anxiety scales and a hybrid questionnaire based on minority and resilience. We conducted interviews, analyzed sentiment with TF-IDF, and constructed classification algorithms. We extracted and validated questionnaire data using median, SD, and Cronbach alpha. Correlation between two independent variables. Chi-square and ANOVA examined transgender depression, anxiety, and age. This link was validated using SVM, XGBoost, Naive Bayes, and Logistic Regression where SVM and Naive Bayes had a better accuracy for depression which was 84.6% and for anxiety SVM gave 76.9%. In the case of NLP linear SVC and Random Forest gave the highest F1 score among others which were 92.30% and 76.92% respectively. Using these methods, we determined that depression is comparatively minimal whereas anxiety is severe. This association can be used to minimize the severity of these mental health problems.