IndoBERT-Based Natural Language Processing for Early Detection of Mental Disorders among Indonesian Gen-Z Students: A Mobile Application Approach with Logistic Regression Baseline

Abstract

Mental health issues have become a growing concern among young adults, while access to professional psychological services remains limited. Most existing digital mental health applications rely mainly on self-report questionnaires and lack the ability to interpret contextual emotional expressions found in user-written text, which reduces their effectiveness for early screening. This study proposes the design and implementation of a mobile-based mental health detection system that integrates contextual natural language processing with interactive assessment features. The system analyzes Indonesian-language textual reflections using an IndoBERT-based classification model and complements the results with a rule-based psychological scoring mechanism derived from questionnaire responses. Logistic Regression with TF–IDF features is employed as a baseline model for comparative evaluation. System performance is assessed using accuracy, precision, recall, and F1-score metrics. Experimental results show that the IndoBERT model outperforms the baseline, achieving an accuracy of 97.79%, compared to 94.17% for Logistic Regression. The proposed system is implemented as a Flutter-based mobile application to improve accessibility to early mental health screening among Indonesian university students. This study integrates two complementary approaches: NLP-based text classification using IndoBERT and rule-based psychological scoring derived from self-report questionnaires.

Keywords

IndoBERT Natural Language Processing; Logistic Regression; Detection Generation Z