DETECTING MENTAL DISORDERS USING EMOTIONAL PATTERNS FROM SOCIAL MEDIA

AbstractMillions of people worldwide suffer from mental disorders that significantly impact their thinking andbehavior. Timely detection of these conditions is both challenging and essential, as early interventioncan provide crucial support before the illness progresses. One promising approach to this challenge ismonitoring how individuals express themselves, particularly in their writing and social mediacommunications. This article examines two computational models designed to capture the presenceand shifts in emotions expressed by social media users. Our analysis utilizes two recent public datasetsfocused on depression and anorexia. The results indicate that the proposed emotional representationseffectively highlight critical information about users affected by these disorders. Moreover, combiningthese models enhances performance, matching the best-reported method for depression and nearlyequaling the top approach for anorexia, with only a 1% difference. Additionally, these models offerpotential for greater interpretability of the result. Keywords: Mental disorders, emotional patterns, machine learning