MLDSS: Customer-Centric Retail Recommendation via Multi-Layered Decision Support System

In the ever-evolving landscape of retail, the need for an advanced recommendation system has become crucial to enhance customer experience and drive sales. This research introduces a novel multilayered recommendation system designed to provide personalized product recommendations by leveraging a combination of machine learning techniques and association rule mining (ARM). To design this system, we made the following contributions: Our first contribution is to generate association rules from the transaction dataset. The second contribution is to extract features from the association rules to have a better input to the proposed model. The third contribution is to design a multilayered recommendation system where, at each layer, we have applied and synchronized different ML techniques to generate customer-centric recommendations. The performance of the proposed system achieves a comparative accuracy of 99.09%, F1 score of 0.95, precision of 0.94, recall of 0.925, and AUC value of 0.987, demonstrating the ability to provide accurate, personalized, and dynamic recommendations. Furthermore, the system is rigorously validated under noise, drift, seasonality, and anomaly simulations, confirming its robustness and fault tolerance for real-world retail deployment. This approach improves the accuracy of recommendations and ensures adaptability to changing customer preferences and market trends, offering a robust solution for modern retail environments.