Depression Detection From Social Media Activity Using Machine Learning

Abstract

In my research project titled "Depression detection from social media activity using machine learning," I investigate the effectiveness of various machine learning algorithms in identifying depressive symptoms from social media data. Throughout the study, I evaluate the performance of Support Vector Machine (SVM), Random Forest, Long Short-Term Memory (LSTM), and Bidirectional LSTM (Bi-LSTM) algorithms. Utilizing a dataset comprising labeled social media posts, I analyze the results obtained from each algorithm. The SVM algorithm achieves an accuracy of 90%, with precision and recall scores of 0.87 and 0.94 for depressed posts, and 0.94 and 0.86 for non-depressed posts, respectively. Meanwhile, Random Forest yields an accuracy of 79%, while LSTM and Bi-LSTM models perform significantly better, achieving accuracies of 97% and 98%, respectively. These deep learning models demonstrate superior performance, as evidenced by their precision, recall, and F1-score metrics. My findings underscore the potential of machine learning, particularly deep learning techniques, in revolutionizing depression detection and mental health monitoring through the analysis of social media activity.