International Journal of Research and Applied Technology (INJURATECH)
Vol. 6 No. 2 (2026): December, 2026 (Issue in Progress)

Optimization of The Retail Supply Chain in Jayapura City through A Hybrid Machine Learning Approach

Erickzak Yusup Nussy (Universitas Komputer Indonesia, Indonesia)
Irfan Dwiguna Sumitra (Universitas Komputer Indonesia, Indonesia)



Article Info

Publish Date
29 Oct 2026

Abstract

The coffee retail supply chain in Jayapura City faces complex operational challenges caused by geographic isolation, heavy reliance on conventional bulk delivery logistics, and infrastructure vulnerabilities such as periodic power outages. Furthermore, nonlinear demand volatility heavily influenced by local socioeconomic dynamics such as community payment cycles often causes conventional single-stage forecasting models to produce severe stock imbalances, stock shortages, or storage cost inflation. To address these limitations, this study develops a two-stage hybrid machine learning framework that combines unsupervised K-Means clustering for daily market trend segmentation with supervised Multi-Output Random Forest regression for simultaneous raw material demand forecasting. This hybrid approach is significantly more suitable for geographically constrained retail environments than conventional models because unsupervised cluster labels act as robust synthetic bridge features, effectively suppressing high-dimensional operational noise and isolating systemic external shocks before the regression stage. Tested on a 365-day operational dataset, K-Means successfully categorized market dynamics into three distinct operational regimes (K=3), achieving an overall average prediction accuracy ( ) of 92.77% with low error metrics (MAE Coffee = 0.20 kg, MAE Milk = 1.19 L, and MAE Cup = 9.18 pcs). Integrated into a Streamlit-based interactive dashboard, the proposed system empowers retail MSMEs in Eastern Indonesia to transition to data-driven inventory control that successfully reduces supply chain risks, storage costs, and raw material spoilage

Copyrights © 2026






Journal Info

Abbrev

injuratech

Publisher

Subject

Civil Engineering, Building, Construction & Architecture Computer Science & IT Control & Systems Engineering Electrical & Electronics Engineering Engineering

Description

INJURATECH cover all topics under the fields of Computer Science, Information system, and Applied Technology. Scope: Computer Based Education Information System Database Systems E-commerce and E-governance Data mining Decision Support System Management Information System Social Media Analytic Data ...