Eksplorasi Potensi Jenis Pekerjaan Berdasarkan Panduan Pemaknaan Hasil Asesmen Bakat Minat Siswa melalui Pembelajaran Mesin dan LLM

Satrianto, Pujo (2026) Eksplorasi Potensi Jenis Pekerjaan Berdasarkan Panduan Pemaknaan Hasil Asesmen Bakat Minat Siswa melalui Pembelajaran Mesin dan LLM. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6022241134-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (15MB) | Request a copy

Abstract

Perubahan kebutuhan kompetensi dan dinamika dunia kerja menuntut layanan bimbingan karier pada satuan pendidikan memanfaatkan data asesmen secara lebih terstruktur dan komunikatif. Penelitian ini membangun alur analitik pemanfaatan data Asesmen Bakat Minat (ABM) untuk menghasilkan narasi eksplorasi potensi pekerjaan siswa. Data berasal dari hasil ABM siswa Madrasah Aliyah di Kabupaten Kutai Kartanegara. Alur penelitian mencakup pembentukan label proksi keselarasan minat-bakat menggunakan Gaussian Mixture Model (GMM), evaluasi kualitas clustering, validasi pakar, klasifikasi XGBoost, pengayaan konteks pekerjaan, penyusunan payload, pembangkitan narasi menggunakan Large Language Model (LLM), dan evaluasi RAGAS. Pelabelan GMM pada 3.070 pasangan siswa-target menghasilkan proporsi label Selaras 54,72%–60,59%. Evaluasi clustering memperoleh Silhouette Score 0,4069–0,4282 dan log-likelihood GMM dua komponen -502,82 sampai -276,90. Validasi pakar menghasilkan agreement 0,827 dan Cohen’s Kappa 0,652. XGBoost memperoleh accuracy agregat 0,9388, F1-score 0,9480, dan ROC-AUC 0,9827. Narasi berhasil dibangkitkan untuk 60 siswa menggunakan GPT-4.1. Evaluasi RAGAS menunjukkan faithfulness 0,512–0,975, context recall 0,749–0,907, context precision 0,868–0,999, dan rubric score 4,917–5,000. Hasil penelitian menunjukkan bahwa data ABM dapat diolah menjadi narasi eksplorasi pekerjaan yang terstruktur, berbasis data, komunikatif, dan mendukung layanan bimbingan karier.
=======================================================================================================================================
Changing competency demands and labor-market dynamics require career guidance services in educational institutions to utilize assessment data in a more structured and communicative manner. This study develops an analytical workflow for utilizing Aptitude and Interest Assessment (ABM) data to generate narratives for exploring students’ potential occupational fields. The data were obtained from ABM results of Madrasah Aliyah students in Kutai Kartanegara Regency. The workflow includes proxy labeling of interest-aptitude alignment using Gaussian Mixture Model (GMM), clustering quality evaluation, expert validation, XGBoost classification, occupational context enrichment, payload construction, narrative generation using a Large Language Model (LLM), and RAGAS evaluation. GMM labeling on 3,070 student-target pairs produced Aligned label proportions of 54.72%–60.59%. Clustering evaluation obtained Silhouette Score values of 0.4069–0.4282 and two-component GMM log-likelihood values ranging from -502.82 to -276.90. Expert validation produced an agreement of 0.827 and Cohen’s Kappa of 0.652. XGBoost achieved an aggregate accuracy of 0.9388, F1-score of 0.9480, and ROC-AUC of 0.9827. Narratives were generated for 60 students using GPT-4.1. RAGAS evaluation showed faithfulness of 0.512–0.975, context recall of 0.749–0.907, context precision of 0.868–0.999, and rubric score of 4.917–5.000. The results indicate that ABM data can be transformed into structured, data-driven, and communicative occupational exploration narratives that support career guidance services.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	ABM, GMM, XGBoost, LLM, rekomendasi pekerjaan, RAGAS, occupational recommendation.
Subjects:	L Education > L Education (General) L Education > LB Theory and practice of education > LB1603 Secondary Education. High schools L Education > LC Special aspects of education > LC5201 Education extension. Adult education. Continuing education Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T58.64 Information resources management
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20101-(S2) Master Thesis
Depositing User:	Pujo Satrianto
Date Deposited:	18 Jul 2026 06:15
Last Modified:	18 Jul 2026 06:15
URI:	http://repository.its.ac.id/id/eprint/135368

Actions (login required)

View Item