Hybrid Markov Regime-Switching and Hurdle modeling of last-mile livestock logistics in Ukraine under systemic volatility

Abstract This study examines the transformation of last-mile logistics in Ukraine’s livestock supply chains under conditions of prolonged systemic volatility and infrastructure disruption. A hybrid analytical framework is proposed that integrates Random Forest Recursive Feature Elimination (RF-RFE) for preliminary predictor screening, a two-state Markov Regime-Switching (MRS) model for identifying latent logistics regimes, and a two-part Hurdle model for separately estimating disruption probability and conditional freight-cost intensity. The empirical analysis uses a verified monthly panel dataset covering 2023–2025 across three contrasting regional logistics nodes - Kyiv, Vinnytsia, and Lviv - and incorporates energy, security, market, transport, and infrastructure-related indicators. The MRS results identify two statistically distinct and persistent freight-cost regimes, with the systemic-crisis regime characterized by a 21.8% higher mean Freight Cost Per Ton and a longer expected duration than the operational-stability regime. The Hurdle estimates show that Security Alerts and Power Outage Hours are positively associated with the probability of logistics disruption, while Power Outage Hours and Fuel Price Volatility are positively associated with conditional freight-cost intensity; regional comparisons further reveal heterogeneous exposure to systemic shocks. Counterfactual technological scenarios further indicate that combining AI-assisted dynamic routing with decentralized autonomous and light electric transport solutions could reduce the estimated sensitivity of freight costs to energy disruptions by approximately 24.2%. The findings demonstrate the value of combining latent-regime identification, sequential disruption modeling, and technological scenario analysis for assessing the resilience of livestock last-mile logistics under persistent infrastructure and security shocks.