Improving Accuracy and Robustness in Depression Detection with Ensemble Learning and Optimization Techniques

Chronic melancholy, lack of interest or enjoyment in activities, changes in diet or weight, abnormalities in sleep patterns, feelings of guilt or unworthiness, difficulty in concentration, and suicide or death thoughts are all symptoms of depression. Depression, if left untreated, can lead to a variety of problems, including an increased risk of self-harm or suicide, substance abuse, physical health issues, and social isolation. In this work we attempt to utilize machine learning techniques to detect a likely depressed social media user based on social media posts and compare the accuracy of built-in models using a variety of comparison methodologies. The machine learning algorithms were fine-tuned on five various optimization techniques.