Conflict aware retail recommendation decision support system

Abstract Retail recommendation systems play a crucial role in enhancing customer engagement and driving sales across modern commerce platforms. Traditional Association Rule Mining (ARM) based approaches often prioritize consistency and completeness in recommendations, overlooking the inherent uncertainty and contradictions in real-world retail data. To address these challenges, this study introduces the Conflict Aware Retail Recommendation Decision Support (CAR2DS) framework, which integrates Paraconsistent Annotated Logic (PAL) with evidential analysis to enable conflict-tolerant and transparent retail recommendations. This framework prioritizes ARM-derived rules based on support-confidence strength, temporal consistency, and reliability indices, ensuring coherent and explainable decision-making under uncertainty. Empirical evaluation reveals that CAR2DS achieves superior performance, attaining an average accuracy of 95.8% and demonstrating mean gains of 23.6% over ARM and 14.3% over FARM . It provides an interpretable decision with validating its transparency and reliability. Additionally, the model demonstrates balanced performance across key metrics, including conflict resolution, stability, and temporal adaptivity. The proposed system establishes a new paradigm for explainable, Conflict Aware retail intelligence.