OPTIMASI PERSEDIAAN BERBASIS MACHINE LEARNING UNTUK MENGURANGI STOCKOUT MENGGUNAKAN ALGORITMA RANDOM FOREST DAN XGBOOST PADA SKU PRIORITAS DI PT. XYZ (Studi Kasus: Industri FMCG)

Daffa Rozano, , (2026) OPTIMASI PERSEDIAAN BERBASIS MACHINE LEARNING UNTUK MENGURANGI STOCKOUT MENGGUNAKAN ALGORITMA RANDOM FOREST DAN XGBOOST PADA SKU PRIORITAS DI PT. XYZ (Studi Kasus: Industri FMCG). Skripsi thesis, Universitas Pembangunan Nasional Veteran Jakarta.

[img] Text
ABSTRAK.pdf
Download (53kB)
[img] Text
AWAL.pdf
Download (1MB)
[img] Text
BAB 1.pdf
Restricted to Repository UPNVJ Only
Download (233kB)
[img] Text
BAB 2.pdf
Restricted to Repository UPNVJ Only
Download (370kB)
[img] Text
BAB 3.pdf
Restricted to Repository UPNVJ Only
Download (248kB)
[img] Text
BAB 4.pdf
Restricted to Repository UPNVJ Only
Download (988kB)
[img] Text
BAB 5.pdf
Download (119kB)
[img] Text
DAFTAR PUSTAKA.pdf
Download (122kB)
[img] Text
RIWAYAT HIDUP.pdf
Restricted to Repository UPNVJ Only
Download (123kB)
[img] Text
LAMPIRAN.pdf
Restricted to Repository UPNVJ Only
Download (583kB)
[img] Text
HASIL PLAGIARISME.pdf
Restricted to Repository staff only
Download (12MB)
[img] Text
ARTIKEL KI.pdf
Restricted to Repository staff only
Download (414kB)

Abstract

PT. XYZ is a company engaged in the Fast-Moving Consumer Goods (FMCG) industry facing significant challenges in managing demand volatility, where the existing system frequently experiences stockouts on priority SKUs due to negative forecasting bias (under-forecast). This study aims to optimize inventory policy by integrating Machine Learning algorithms to improve product availability reliability. Based on the Syntetos-Boylan Classification (SBC), the demand characteristics of priority SKUs are identified as erratic, indicating high variability. The methodology compares Random Forest and XGBoost algorithms, where the Root Mean Squared Error (RMSE) of the best model is integrated into dynamic Safety Stock calculations, incorporating a 22-day protection period to accommodate the fixed production cycle constraint. The results indicate that the XGBoost algorithm with an 80:20 data split scenario is the best-performing model. Despite having a higher RMSE compared to the existing method, this model possesses a strategic advantage in the form of safety bias capable of mitigating stock scarcity risks. The simulation of the proposed inventory policy proved effective in drastically reducing the stockout rate from 64.00% to 0.62%, and increasing the Service Level from 36.00% to 99.38%. Furthermore, capacity validation using the Historical Peak Benchmark approach demonstrates that the proposed system is operationally feasible, utilizing only 18.4% of the historical maximum storage capacity in the region with the highest volume.

Item Type: Thesis (Skripsi)
Additional Information: [No.Panggil: 2110312063] [Pembimbing: Yulizar Widiatama] [Penguji 1: Alina Cynthia Dewi] [Penguji 2: M. Rachman Waluyo]
Uncontrolled Keywords: Machine Learning, XGBoost, Inventory Optimization, Dynamic Safety Stock, FMCG.
Subjects: T Technology > T Technology (General)
T Technology > TA Engineering (General). Civil engineering (General)
Divisions: Fakultas Teknik > Program Studi Teknik Industri (S1)
Depositing User: DAFFA ROZANO
Date Deposited: 05 Mar 2026 06:18
Last Modified: 05 Mar 2026 06:18
URI: http://repository.upnvj.ac.id/id/eprint/41908

Actions (login required)

View Item View Item