DEEP RISK INSIGHT: REAL-TIME SUPPLY CHAIN DISRUPTION PREDICTION USING LSTM NETWORKS

Abstract

Supply chain disruptions present critical challenges across global industries, affecting production efficiency, cost control, and customer satisfaction. Traditional risk management approaches often depend on historical data, rule-based frameworks, and expert judgment, which fall short in anticipating disruptions driven by dynamic and unpredictable factors like geopolitical instability, climate change, and fluctuating demand. With the rise of machine learning, predictive capabilities have improved, yet conventional classification models struggle to effectively capture the temporal nature of supply chain data. This study introduces a deep learning methodology that leverages Long Short-Term Memory (LSTM) networks to enhance disruption prediction by modeling sequential patterns within supply chain activities. LSTM's strength in retaining long-term dependencies is further combined with a Random Forest Classifier (RFC) to enhance decision accuracy based on temporal risk trends. In comparison to traditional models such as decision trees and logistic regression, the LSTM-based framework adapts to real-time data and delivers improved prediction accuracy. Despite these advancements, current systems still face challenges, including data imbalance, limited interpretability, and high computational demands. To overcome these, the proposed approach incorporates SMOTE (Synthetic Minority Over-sampling Technique) to balance class distribution, applies feature selection for better model efficiency, and utilizes a hybrid deep learning model for more reliable predictions. The results show that the proposed LSTM-RFC model significantly outperforms traditional techniques in early detection of supply chain disruptions, enabling organizations to implement proactive risk mitigation. This research contributes to reducing operational vulnerabilities, strengthening supply chain resilience, and improving strategic decisionmaking across sectors such as manufacturing, logistics, and retail