Explainable AI for Operational Risk Management in Logistics and Procurement Systems

Logistics and procurement of management operations affect the stability of supply chains of organizations, financial loss, and continuity of services. Nevertheless, traditional predictive models tend to be black boxes, meaning that one has a limited understanding of the logic behind based risk predictions hence decreasing the stakeholder and limiting effective decision-making. To counter this problem, in this research, the aim and goal are to develop an explainable AI system that is integrated with the accuracy and interpretability of operational risk detection. This research is new in the sense that a hybrid model, known as RiskSHAP-AttentivePathNet was designed, which combines AttentivePathNet, an attention-based sequential learner model, with SHAP (SHapley Additive Explanations) to offer both sequential and post-hoc attribution to features. The suggested structure could show a high accuracy level in prediction of 98%, which is by far better than traditional models of LSTM, Random Forest, and XGBoost, which demonstrated an average of 86% to 92% accent level. In addition to the high quality of predictiveness, the model suggests in-depth explanations and SHAP based visualizations, attention path tracking, and can explain each risk prediction. To sum up, there is a bright future of RiskSHAP-AttentivePathNet operating as a user-friendly, efficient, and transparent decision support tool in operational risk management in a complicated logistics/procurement environment.

Hali tarjima qilinmagan