A Predicting Depression Model from Social Media Images using Machine Learning Technique

The rate of depression in Thailand has been steadily increasing, with the majority of cases going untreated. As a result, expressions of depression often appear on social media. This study aims to develop a predicting depression model from social media images using machine learning technique. Data were collected from Twitter users, including images and results from a Patient Health Questionnaire-9 (PHQ-9), totaling 1,131 images. These were categorized into four groups: 423 images of individuals without depression, 525 images of individuals with mild depression, 134 images of individuals with moderate depression, and 69 images of individuals with severe depression. A convolutional neural network was applied in the study. The experimental results showed that the predictive model for depression based on images from social networks using machine learning techniques achieved the accuracy rate is 81.16%, the precision rate is 81.88%, the recall rate is 80.81% and the overall efficiency rate is 81.02%.