Deteksi Dini Gangguan Mental Menggunakan Klasifikasi Teks Berbasis Machine Learning

Ahnaf, Muhammad Firuz (2025) Deteksi Dini Gangguan Mental Menggunakan Klasifikasi Teks Berbasis Machine Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2043211044-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Masalah kesehatan mental pada remaja menjadi perhatian serius di Indonesia, terutama setelah hasil survei I-NAMHS 2022 menunjukkan bahwa sekitar 5,5% remaja berusia 10–17 tahun mengalami gangguan mental. Kondisi ini menuntut upaya deteksi dini yang lebih efisien dan menjangkau kelompok usia muda. Penelitian ini bertujuan untuk mengembangkan sistem klasifikasi gangguan mental berbasis teks dari media sosial, khususnya Twitter, dengan pendekatan machine learning. Langkah pertama dalam penelitian ini adalah mengumpulkan dan memberi label secara manual sebanyak 3.000 tweet berdasarkan dua tahap klasifikasi: relevansi dan kategori (Terindikasi, Penderita, Selfdiagnosed, Penyintas). Proses pelabelan kemudian dilanjutkan secara iteratif dengan semi-supervised learning. Model klasifikasi dibangun menggunakan tiga algoritma machine learning: Support Vector Machine, Random Forest, dan Naive Bayes. Hasil terbaik untuk klasifikasi relevansi diperoleh dari SVM dengan akurasi 97,08%, presisi 96,93%, recall 97,08%, dan F1-score 96,94%. Untuk klasifikasi kategori, performa terbaik dicapai oleh Random Forest dengan akurasi 88,00%, presisi 87,82%, recall 88,00%, dan F1-score 86,70%. Analisis lebih lanjut menunjukkan bahwa model relevansi masih menghadapi tantangan dalam membedakan narasi pengalaman pribadi dengan referensi atau saran, sedangkan model kategori belum sepenuhnya menangkap konteks waktu, subjek pengalaman, dan status diagnosis. Selain itu, dikembangkan aplikasi web yang mampu mengklasifikasikan teks terkait kesehatan mental secara real-time dan memberikan umpan balik, termasuk rekomendasi untuk konsultasi profesional apabila terdeteksi gejala gangguan mental. Penelitian ini dapat dikembangkan lebih lanjut dengan mengadopsi teknik representasi teks berbasis contextual embeddings.

=====================================================================================================================================
Mental health problems in adolescents is a serious concern in Indonesia, especially after the results of the 2022 I-NAMHS survey showed that around 5.5% of adolescents aged 10-17 experienced mental disorders. This study aims to develop a text-based mental disorder classification system from social media, specifically Twitter, with a machine learning approach. The first step in this research is to manually collect and label 3,000 tweets based on two classification stages: relevance and category (Help-seeking, Clinically Diagnosed, Selfdiagnosed, Survivor). The labeling process was then continued iteratively with semi-supervised learning. The classification model was built using three machine learning algorithms: Support Vector Machine, Random Forest, and Naive Bayes. The best results for relevance classification were obtained from SVM with 97.08% accuracy, 96.93% precision, 97.08% recall, and 96.94% F1-score. For category classification, the best performance was achieved by Random Forest with 88.00% accuracy, 87.82% precision, 88.00% recall, and 86.70% F1-score. Further analysis shows that the relevance model still faces challenges in distinguishing personal experience narratives from references or suggestions, while the category model has not fully captured the context of time, experience subject, and diagnosis status. In addition, a web application was developed that is able to classify mental health-related texts in real-time and provide feedback, including recommendations for professional consultation if symptoms of mental disorders are detected. This research can be further developed by adopting text representation techniques based on contextual embeddings and expanding data coverage to other social media platforms to improve coverage and detection accuracy.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Gangguan Mental, Klasifikasi Teks, Naive Bayes, Random Forest, SVM, Twitter, Mental Disorder, Text Classification.
Subjects:	H Social Sciences > HA Statistics Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA279.5 Bayesian statistical decision theory. Q Science > QA Mathematics > QA76.6 Computer programming. Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science)
Divisions:	Faculty of Vocational > 49501-Business Statistics
Depositing User:	Muhammad Fairuz Ahnaf
Date Deposited:	09 Jul 2025 08:24
Last Modified:	09 Jul 2025 08:24
URI:	http://repository.its.ac.id/id/eprint/119450

Actions (login required)

View Item