Datasets

The rise in social media usage has made it a crucial platform for analyzing behaviors, including detecting signs of depression through posts and comments. This study investigates the effectiveness of deep learning models like XLNET, DistilBERT, and ALBERT, along with traditional machine learning models, in identifying depression-related content on social media, using a Reddit dataset. Among the models, XLNET with text summarization achieved the highest accuracy of 63.33%. Additionally, the study explores task-incremental learning in a continual learning setting using Elastic Weight Consolidation, demonstrating improved performance by preserving critical weights and balancing knowledge.