Pengembangan Model Peramalan Kebutuhan Suku Cadang Intermittent dan Lumpy Menggunakan SVR: Studi Kasus Industri Perkeretaapian (PT XYZ)

Afghoni, Syayid Al (2026) Pengembangan Model Peramalan Kebutuhan Suku Cadang Intermittent dan Lumpy Menggunakan SVR: Studi Kasus Industri Perkeretaapian (PT XYZ). Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032241001-Master_Thesis.pdf Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Workshop perawatan industri perkeretaapian menghadapi permasalahan dalam meramalkan kebutuhan suku cadang, 89% item memiliki pola permintaan intermittent dan lumpy. Penelitian ini bertujuan mengembangkan dan mengevaluasi model peramalan berbasis Support Vector Regression (SVR) dengan kernel RBF melalui dua skenario, SVR-1 menggunakan variabel jadwal perawatan P24 dan P48, serta SVR-2 yang mengintegrasikan tambahan variabel klasifikasi umur operasional kereta (<15, 15–30, >30 tahun). Sepuluh item suku cadang Vital dan Essential dipilih melalui purposive sampling berbasis analisis ADI-CV² dan VED. Hyperparameter dioptimasi menggunakan Randomized Search dengan Time Series Cross-Validation. Model dievaluasi terhadap tiga metode pembanding, metode eksisting, Croston, dan Random Forest. Hasil menunjukkan SVR-2 mencapai rata-rata MASE terbaik sebesar 0,661 unggul 30,2% atas metode eksisting (0,947) dan konsisten terbaik pada 8 dari 10 item. Dari sisi total biaya persediaan, SVR-2 menghasilkan penghematan sebesar 38,5% (Rp 452,8 juta) dibandingkan metode eksisting. Variabel umur kereta terbukti meningkatkan akurasi secara empiris terutama pada komponen wear-sensitive seperti komponen pengereman dan mekanik. Implementasi model SVR ini diharapkan dapat menjadi landasan bagi manajemen PT XYZ dalam menyusun strategi pengadaan suku cadang yang lebih efisien, meminimalkan risiko stockout dan overstock, serta memperkuat pengambilan keputusan berbasis data.
========================================================================================================================================
A railway industry maintenance workshop faces challenges in spare parts demand forecasting, 89% items exhibit intermittent and lumpy demand patterns. This study aims to develop and evaluate a Support Vector Regression (SVR) forecasting model using an RBF kernel across two scenarios: SVR-1, which uses the P24 and P48 maintenance schedule variables, and SVR-2, which integrates an additional variable classifying train operational age (<15, 15–30, >30 years). Ten Vital and Essential spare parts were selected through purposive sampling based on ADI-CV² and VED analysis. Hyperparameters were optimized using Randomized Search with Time Series Cross-Validation. The model was evaluated against three comparison methods: the existing method, Croston, and Random Forest. The results show that SVR-2 achieved the best average MASE of 0.661, outperforming the existing method (0.947) by 30.2% and consistently performing best on 8 out of 10 items. In terms of total inventory costs, SVR-2 generated savings of 38.5% (Rp 452.8 million) compared to the existing method. The train age variable has been empirically shown to improve accuracy, particularly for wear-sensitive components such as braking and mechanics systems. The implementation of this SVR model is expected to serve as a foundation for PT XYZ’s management in developing more efficient spare parts procurement strategies, minimizing the risks of stockouts and overstocking, and strengthening data-driven decision-making.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Intermittent, Lumpy, Peramalan Permintaan Suku Cadang, Support Vector Regression, Spare Parts Demand Forecasting
Subjects:	H Social Sciences > HD Industries. Land use. Labor > HD38.5 Business logistics--Cost effectiveness. Supply chain management. ERP Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Syayid Al Afghoni
Date Deposited:	05 Jul 2026 07:54
Last Modified:	05 Jul 2026 07:54
URI:	http://repository.its.ac.id/id/eprint/134295

Actions (login required)

View Item