A Multi-store Recommendation and Route Optimization System for Enhanced Grocery Shopping

Grocery shopping often suffers from inefficiencies caused by price variations, inaccurate inventories, and the absence of personalized recommendations. This paper presents a system that optimizes both store selection and travel routes for a user’s shopping list. A parallelized branch-and-bound algorithm identifies cost-efficient store combinations, while an Adaptive Genetic Algorithm (AGA) with A* search computes traffic-aware routes. A collaborative filtering model using Singular Value Decomposition (SVD) predicts future purchases. The architecture integrates caching and routing servers to support real-time scalability. Experiments on a benchmark dataset derived from local retailer surveys show the parallel branch-and-bound approach achieves strong balance between optimality and efficiency, with route planning reducing travel cost by 10–20% and SVD maintaining high F1-scores under sparsity. Qualitative surveys informed design choices such as limiting recommendations to three or four stores, demonstrating that integrating optimization and personalization can substantially improve the grocery shopping experience.