Text generation identifikasi kerusakan pada kendaraan bermotor mengggunakan indogpt

Damayanti, Salsabila (2025) Text generation identifikasi kerusakan pada kendaraan bermotor mengggunakan indogpt. Sarjana thesis, UIN Sunan Gunung Djati Bandung.

Preview	Text 1_cover.pdf Download (2MB) | Preview
Preview	Text 2_abstrak.pdf Download (549kB) | Preview
Preview	Text 3_skbebasplagiarism.pdf Download (447kB) | Preview
Preview	Text 4_daftarisi.pdf Download (113kB) | Preview
Preview	Text 5_bab1.pdf Download (227kB) | Preview
	Text 6_bab2.pdf Restricted to Registered users only Download (401kB) | Request a copy
	Text 7_bab3.pdf Restricted to Registered users only Download (279kB) | Request a copy
	Text 8_bab4.pdf Restricted to Registered users only Download (653kB) | Request a copy
	Text 9_bab5.pdf Restricted to Registered users only Download (131kB) | Request a copy
	Text 10_daftarpustaka.pdf Restricted to Registered users only Download (183kB) | Request a copy

Abstract

INDONESIA: Penelitian ini berfokus pada pengembangan sistem identifikasi kerusakan kendaraan bermotor berbasis text generation dengan memanfaatkan model IndoGPT. Tujuan utamanya adalah mengevaluasi kemampuan model dalam menghasilkan respons berbentuk solusi secara otomatis dari masukan berupa gejala kerusakan kendaraan. Proses pengembangan mengikuti tahapan metodologi CRISP-DM, mulai dari pemahaman permasalahan hingga tahap evaluasi model. Tiga model IndoGPT digunakan dalam proses fine-tuning, yaitu indobenchmark/indogpt, w11wo/indo-gpt2-small, dan cahya/gpt2-small-indonesian-522M, untuk menghasilkan teks solusi dari input gejala. Evaluasi dilakukan melalui pengamatan terhadap performa pelatihan serta pengujian berbasis prompt yang dibandingkan dengan jawaban referensi guna menilai tingkat kesesuaian hasil keluaran model. Hasil menunjukkan bahwa model indobenchmark/indogpt memberikan performa paling konsisten, dengan nilai loss terendah dan tingkat akurasi jawaban tertinggi dibandingkan dua model lainnya. Penelitian ini berkontribusi dalam penerapan teknologi NLP untuk sektor otomotif, khususnya dalam mendukung proses identifikasi kerusakan kendaraan secara otomatis pada tahap awal. ENGLISH: This research focuses on developing a motor vehicle damage identification system based on text generation using the IndoGPT model. The main objective is to evaluate the model's ability to automatically generate solution-based responses from input in the form of vehicle damage symptoms. The development process follows the CRISP-DM methodology, from problem understanding to model evaluation. Three IndoGPT models were used in the fine-tuning process: indobenchmark/indogpt, w11wo/indo-gpt2-small, and cahya/gpt2-small-indonesian-522M, to generate solution texts from symptom inputs. Evaluation was conducted by observing training performance and prompt-based testing, which were compared with reference answers to assess the level of consistency in the model's output. The results show that the indobenchmark/indogpt model provides the most consistent performance, with the lowest loss value and the highest answer accuracy rate compared to the other two models. This research contributes to the application of NLP technology in the automotive sector, particularly in supporting the process of automatically identifying vehicle damage at an early stage.

Actions (login required)

View Item