Empirical Juridical-Digital Twin Integrating Machine-Learning Demand Forecasts, Hygiene-Compliance Rules, and Multi-Objective Routing Optimization for Culinary Micro-Enterprises in Pesisir Barat

Background: Culinary micro-enterprises in Pesisir Barat operate under highly volatile demand patterns, stringent hygiene regulations, and elevated transport costs due to the region's fragmented road-network and maritime exposure. Objectives: To construct a legally enforceable digital-twin framework that simultaneously optimises demand forecasting, food-quality preservation, and vehicle routing while embedding contractual risk-allocation mechanisms compliant with the Indonesian Civil Code (KUHPerdata). Methods: A gradient-boosted regression tree (GBRT) was trained on 1,200 hourly observations from 3,200 enterprises (2023-2024) to predict demand with a mean absolute percentage error of 6.3 %. Predicted demand drives a time-integrated Arrhenius model that quantifies spoilage as a function of dynamic temperature exposure; model parameters were calibrated using 4,800 temperature-log entries from on-board sensors. A multi-objective mixed-integer linear program (MO-MILP) minimises weighted fuel consumption, spoilage loss, and contractual penalty costs while maximising service-level adherence. The MO-MILP incorporates (i) force-majeure clauses derived from BMKG sea-state ≥ 2.5 m or official disaster declarations, (ii) graduated penalty schedules for temperature deviation (per García and Hernández, 2022), and (iii) liquidated-damages provisions consistent with Articles 1234-1245 KUHPerdata. Real-time GPS and temperature streams update the digital-twin state-space with latency under 5 s; a blockchain-based smart-contract layer automates penalty assessment and payment settlement. Published in Journal of Coastal Frontiers and Applied Development (JCFAD). CERN Zenodo Record: https://doi.org/10.5281/zenodo.24073588. Publisher: PT CHAIN ASIA CONECT Research and Development Division.