Development and Implementation of a Web Application for Monitoring Social Media Use in Adolescents using Predictive Models Based on Machine Learning

The rising use of social media among adolescents has led to a concerning increase in mental health issues, such as depression and anxiety, highlighting the need for tools that can identify and address these risks early on. This paper presents a web application that employs Random Forest models, combining machine learning with cognitive-behavioral approaches to diagnose social media addiction risks in real-time among adolescents, facilitating personalized interventions for parents and health professionals. Additionally, social media addiction, characterized by Network Obsession (OR), Lack of Personal Control (FCP), and Excessive Use (UE), has shown negative impacts on adolescents' mental health. To address this issue, the dataset, comprising over 10,000 records, was gathered through an ARS questionnaire and real-time monitoring data on social media behavior, including connection hours and types of content consumed. Experiments conducted on a sample of 120 students from Colegio Trini de Carabayllo, out of a population of 450, demonstrated outstanding results: the FCP model achieved 100% accuracy and an F1-score of 100%, effectively detecting lack of personal control; the OR model attained 98.29% accuracy and an F1-score of 99.04%, highly effectively identifying cases of network obsession; and the UE model reached 93.93% accuracy and an F1-score of 96.10%, detecting excessive use patterns. Finally, the recommendations included setting time limits on social media and cognitive-behavioral therapy through the Mentalu application.