SXI++: An AI‐Driven Framework for Enhancing Supply Chain Efficiency and Optimization

Manuscript submitted October 17, 2025; accepted October 29, 2025; published December 12, 2025

Abstract—Supply chain disruptions, including backorders and shipment delays, significantly impact operational efficiency, customer satisfaction, and financial performance. Traditional forecasting methods often struggle with imbalanced data, dynamic demand fluctuations, and logistical uncertainties, making it challenging to predict and mitigate disruptions effectively. Machine Learning (ML) models have emerged as promising solutions for supply chain risk management, yet many suffer from limitations in accuracy, adaptability, and interpretability. In this study, we propose the SXI++ framework as an advanced predictive solution, offering high accuracy, actionable insights, and enhanced supply chain decision-making capabilities. The SXI++ algorithm is a dynamic scoring system that converts complex, multidimensional supply chain data into actionable insights by leveraging composite scores/weights from multiple machine learning and deep learning models. By integrating a proprietary deep neural network, the model refines its predictive correlations with supply chain outcomes, achieving better accuracy while providing interpretable pathways to optimize logistics and inventory management. This study utilizes a dataset comprising 250,000 records for backorder and 30,101 records for shipment delay analysis. Key features include supplier performance, forecasted demand, past sales trends, shipment tracking data, geographical factors, and courier performance metrics. Missing values were imputed using advanced statistical techniques to ensure data integrity. The SXI++ framework employs iterative calibration, dynamic weight adjustments, and deep neural network modeling to enhance predictive accuracy. Performance metrics, including accuracy, precision, and AUC, were calculated to evaluate the model’s effectiveness. Additionally, decision tree analyses were conducted to provide interpretable pathways for reducing backorders and shipment delays, identifying critical operational factors and targeted optimization strategies. The SXI++ algorithm demonstrated superior predictive capabilities, achieving an accuracy of 99.48% for backorder prediction and 99.60% for shipment delay classification, significantly outperforming traditional ML models. Precision and AUC scores reached 97.10% and 0.99, respectively, underscoring the model's reliability. The study established a strong correlation between optimized SXI scores and improved supply chain performance, with a 30.05% reduction in SXI scores resulting in an 87.71% decrease in backorders, while a 9.23% increase in SXI scores led to an 85.13% reduction in shipment delays. Decision tree analyses identified key factors influencing supply chain inefficiencies, such as supplier lead times, forecast accuracy, and transportation constraints, providing actionable recommendations for improvement. The SXI++ platform predicts supply chain risks with nearperfect accuracy and precision using deep learning architectures and iterative calibration. Its ability to direct targeted actions and improve supply chain operations is highlighted by its substantial correlation with reduced backorders and shipment delays. In contrast to conventional models, supply chain managers may proactively reduce risks, optimize logistics, and boost overall efficiency with SXI++’s interpretable decision tree paths.

keywords—backorders, shipment delays, stockouts, logistic and transportations, supply chain optimization, SXI++, Deep Neural Networks (DNN), predictive modelling

Cite: Prasoon Jha, Prashant Yadav, Mahesh Banavar, Srinivas Kilambi,"SXI++: An AI‐Driven Framework for Enhancing Supply Chain Efficiency and Optimization," Journal of Advances in Artificial Intelligence, vol. 3, no. 4, pp. 297-314, 2025. doi: 10.18178/JAAI.2025.3.4.297-314

Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).