Prediksi Risiko Kesehatan Mental Berdasarkan Pola Penggunaan Perangkat Digital Menggunakan Algoritma Logistic Regression

Abstract

The rapid development of digital technology has transformed human interaction patterns but also raised concerns regarding mental health risks caused by excessive digital device usage. This study aims to predict mental health risk based on digital device usage patterns using the Logistic Regression method. The dataset consists of 3,500 records with 24 variables, including digital behavior (device usage duration, number of notifications, and phone unlock frequency) and psychological indicators (stress level, GAD-7 anxiety score, and PHQ-9 depression score). The research process follows the CRISP-DM framework comprising business understanding, data understanding, data preparation, modeling, evaluation, and deployment. The results show that device hours per day, sleep quality, and anxiety score significantly influence mental health risk, achieving a model accuracy of 86% and a ROC-AUC value of 0.89. These findings demonstrate that the Logistic Regression method is effective in identifying individuals at high risk of mental health disorders and has the potential to support early detection through digital technology–based solutions in society.