DLDSS: Dual-Layered Decision Support System for Retail Recommendation

Predictive decision-making has emerged in the dynamic retail landscape to maximize sales and enhance customer satisfaction. This study introduces an Advanced Dual-Layered Predictive Decision-Making Approach to address challenges in retail decision-making by leveraging analytics and machine learning techniques. The proposed model integrates Association Rule Mining (ARM) for uncovering hidden patterns, Feature Engineering to enrich data representation, and a Dual-Layered recommendation system. The first layer generates personalized recommendations and prioritizes the recommendations, while the second layer generates outliers. To ensure robust decision-making, the model incorporates Isolation Forest for outlier detection, enabling the identification and exclusion of anomalous data points. Precision, recall, F1 score, and mean absolute error (MAE) metrics evaluate the system’s performance. Experimental results in a retail dataset demonstrate significant improvements in predictive accuracy, with an F1 score of 0.92 and an MAE reduction of 15% compared to baseline models. This approach provides a scalable and adaptive framework for retail decision-making, enabling real-time optimization of inventory, promotions, and customer engagement strategies. Integrating ARM and machine learning ensures both interpretability and predictive power, making it a valuable tool for modern retail analytics.