Detection of Depression from Social Media Using Deep Learning Approach

Abstract

Depression is a psychological health problem or disorder that is incredibly common today. It's serious and contains a growing variety of issues that have an effect on one's means of life and interferes with the functioning of surroundings. Moreover, it's several prejudicious impacts on society in addition to the country, that result in deteriorating the society. Taking into account the quick ascension of several social media platforms, as a result, It has an impact on society and a person's psychological environment since it serves as a stage for depressed people to express their sentiments and emotions, as well as analyze their actions through social media platforms. This research's main purpose is to see if it's possible to forecast a user's mental state using data to characterize them as depressed or not depressed by utilizing data from Twitter. The linguistic context of the theme narratives is assessed using deep learning models based on the thematic content of the user's tweet. Two Deep Learning architectures namely, CNN and LSTM will be combined in the planned model as a hybrid CNN-LSTM model, which once optimized, achieves a 97% accuracy on a benchmark depression dataset with tweets. Our model improves predictive performance for early detection of depression, according to experimental data based on a variety of performance indicators.