This study aimed to develop innovative media using artificial intelligence (AI) to predict depression from social media texts. Utilizing natural language processing techniques and machine learning algorithms, the developed model focused on analyzing words, messages, or images that tend to indicate negative, neutral, or positive emotional states. Data were sourced from social media platforms such as Facebook, X (Twitter), and Instagram. These data were processed through sentiment analysis to categorize the messages into three levels of severity: severe, moderate, and mild. The predictive data were validated by experts from various fields, including computer science and information technology, psychology, and media studies, to ensure accuracy and reduce potential biases in the model. The findings indicated that the naïve Bayes classifier demonstrates the highest efficiency in predicting negative sentiment, achieving an average accuracy of 88.17% for training sets and 85.00% for testing sets. The F1-measure for negative messages reached 76.70%, reflecting the model’s strong capability to detect depression-related text. However, the model encountered limitations in predicting neutral messages, particularly those involving sarcasm or metaphorical expressions. This research highlights the potential of AI applications in predicting and managing depression on social media. The system can provide timely alerts to individuals at risk and recommend seeking professional psychiatric consultation, offering an effective approach for early detection and intervention in depressive disorders.
Boonsomthop et al. (2026) studied this question.