PREDICTION OF LONELINESS LEVELS FROM SOCIAL MEDIA PROFILES USING THE RANDOM FOREST ALGORITHM

Abstract

In today's fast evolving digital world, social media has become a crucial part of modern life, especially for younger generations like Generation Z at SMK Tunas Media. Born at the start of the 21st century, they are the first generation to grow up with constant access to the internet, technology, and social media. However, recent studies have shown that, despite being raised in a highly connected world, Generation Z is experiencing higher levels of loneliness compared to previous generations. One of the main reasons for this is excessive use of social media, which leads them to prefer digital communication over face to face interactions. Around 73% of them report feeling isolated. This sense of isolation is impacting their mental health, with 91% feeling physically or emotionally stressed, and 68% worrying about the future. Because of this, it’s essential to detect early signs of loneliness in students at SMK Tunas Media who might be at risk. Machine learning, specifically using the Random Forest algorithm, can be used to predict students' loneliness levels based on their social media profiles. This algorithm can analyze data from social media activity, such as behavior patterns, the content shared, social interactions, emotional expressions, and network connections. Additionally, integrating the UCLA Loneliness Scale a widely used tool to measure loneliness into the machine learning model's predictions provides further validation in assessing loneliness. This study shows significant improvements in the classification model's performance after applying the Random Forest method optimized with Optuna. In this study, the model predicted loneliness levels based on social media profiles and achieved an impressive 90% accuracy. This suggests that the Random Forest model optimized with Optuna holds great promise for effectively classifying students' loneliness levels, whether low or high, in Generation Z.