Identification of Anorexia Nervosa Through Indonesian-Language Posts on X Social Media Using the Gated Recurrent Unit Algorithm

Anorexia Nervosa (AN) is a complex Eating Disorder (ED) influenced by biological, psychological, and environmental factors. Individuals affected often have a distorted perception of their body image, leading them to engage in extreme behaviors to achieve or maintain an abnormally low body weight. With a high relapse rate and mortality, effective treatment remains a significant challenge. Early identification plays a critical role in increasing the chances of recovery. However, most existing research is conducted in English, creating a gap in understanding this disorder within the Indonesian language context. This study analyzes 2,000 Indonesian-language posts from the social media platform X using a Gated Recurrent Unit (GRU) model to identify potential indicators of AN. By splitting the dataset into 70% training and 30% testing and training the model over 70 epochs, the best classification accuracy achieved is 85.50%. The data ratio and number of epochs significantly impact the model's accuracy. The accuracy percentage shows that the model successfully identified Anorexia Nervosa quite well. The results of this identification can help psychiatrists in early identification of their patients' AN tendencies.