Detecting Depression in Reddit Posts using Hybrid Deep Learning Model LSTM-CNN

Abstract

The detection of depression is a critical issue for human well-being. Previous research has shown us that online detection is successful in social media, allowing for proactive intervention for depressed users. It is a serious psychological disorder and it takes hold of more than 300 million people across the globe. A person who is depressed experience anxiety and low self-esteem in their everyday life, which affects their relationships with their family and friends, and can lead to various diseases and, in the most extreme scenario, suicide. With the rise of social media, the majority of individuals now use it to express their emotions, feelings, and thoughts. If a person's depression can be discovered early by analyzing their post, then essential efforts can be taken to save them from depression-related disorders or, in the best scenario, from suicide. The main goal of our work is to inspect Reddit user posts to see whether any factors suggest depression attitudes among relevant internet users. We use sentiment examination and Machine Learning (ML) techniques to train the ML model and assess the efficacy of our suggested strategy for this goal. A lexicon of phrases that are more common in depressed accounts is identified. In this study, we have combined Long Short-Term Memory (LSTM) and Convolution Neural Network (CNN) to build a hybrid model that can predict depression by evaluating user textual messages.