Text

PENGEMBANGAN MODEL CHATBOT GENERATIVE AI KESEHATAN MENTAL MAHASISWA DALAM MENDUKUNG SUSTAINABLE DEVELOPMENT GOALS (SDGs) BERBASIS FINE-TUNING LLM

---

Student mental health in academic settings is a critical factor in determining the success of higher education. Various studies have demonstrated the high prevalence of mental disorders, such as anxiety and depression, which reach 41% and 33.6% globally, respectively. The extremely low ratio of professional psychologists, especially in Indonesia, creates a significant service gap, requiring adaptive and measurable digital support solutions. Although Large Language Models (LLMs) offer great potential, conventional chatbots are still limited in terms of empathy, contextual relevance, and adaptation to local communication diversity, especially the phenomenon of code-mixing that commonly occurs in student interactions. This study proposes the development of an Empathetic Conversation Chatbot based on LLMs to address service gaps and language style barriers. This model is implemented through a fine-tuning process on the open-source Qwen 2.5-7B model using the Low-Rank Adaptation (LoRA) technique. This strategy was chosen to ensure optimal computational resource efficiency and facilitate domain-specific adjustments. The training data utilised a dataset containing 15,000 anonymised empathetic conversations, transcripts from counsellor-client sessions, and enriched with code-mixing structures. The fine-tuning process was conducted over 9.5 hours using NVIDIA RTX 4500 computational resources. The resulting model demonstrated significant improvements in empathetic responses and code-mixing handling compared to the unadjusted baseline model. Response quality was evaluated using language style metrics, yielding a BLEU score of 0.20 and a BERT score of 0.774. These figures exceed the results of the baseline Qwen 2.5-7B model, which achieved a BLEU score of 0.15 and a BERT score of 0.73, confirming the effectiveness of contextual adaptation. The stateless architecture design, equipped with a short-term memory mechanism, ensures privacy and conversation continuity. This development offers a scalable, parameter-efficient, responsive, and proven accurate approach to student mental health support.

---

Availability

Detail Information

Series Title

-

Call Number

T1881862025

Publisher

Palembang : Prodi Doktor Ilmu Teknik, Fakultas Teknik Universitas Sriwijaya., 2025

Collation

xiii, 195 hlm.; ilus.; tab, 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

610.285 07

Content Type

Text

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Prodi Doktor Ilmu Teknik
Penerapan Teknologi Komputer

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment