Predicting Emotional Well-Being of Gen Z from Social Media Activity Using LSTM: A Deep Learning Approach to Digital Behavior Analysis

Closed

M. Gibran Rifka Pramudia, A'Imatul Mustafidah, Bilqis Jauharah Ummah, Bima Setyo Nugroho, Hujjatullah Fazlurrahman, Dhiyan Septa Wihara, Fajar Wahyudi Rahman, Achmad Kautsar, Nadia Asandimitra Haryono

2025 Proceeding - 2025 International Conference on Information Technology Research and Innovation: Harnessing Intelligent Machines for Sustainable Development: Aligning AI with the SDGs, ICITRI 2025 Conference paper Cited by 1

Abstract

This study investigates the use of Long Short-Term Memory (LSTM), a deep learning model, to predict emotional well-being based on social media activity. Trained on user-generated content from platforms like Instagram, Twitter, and TikTok, the model classifies six emotion categories. Results show strong performance in detecting dominant emotions (happy, neutral), though performance declines for minority classes (fear, disgust) due to data imbalance and overlapping features. These findings highlight the potential of AI in monitoring mental health and reinforce social media's role as a reflection of emotional patterns, particularly among Generation Z. © 2025 IEEE.

Affiliations

Universitas Negeri Surabaya, Faculty of Economics and Business, Departement of Digital Business, Surabaya, Indonesia; Universitas Negeri Surabaya, Faculty of Economics and Business, Departement of Economics, Surabaya, Indonesia; Universitas Negeri Surabaya, Faculty of Economics and Business, Departement of Management, Surabaya, Indonesia

View full text (DOI) →