Digital Economy-Driven Collaborative Scheduling Optimization for E-commerce Fulfillment Using Enhanced K-medoids Clustering with BWP and Local Search Integration

Abstract

In the era of digital economy, the new retail e-commerce industry faces increasingly personalized and diversified consumer demands that require optimized collaborative scheduling to complete orders. An enhanced K-medoids clustering algorithm that integrates a Balanced Weighted Performance (BWP) metric and a Large Neighborhood Search (LNS) mechanism is proposed to address the inefficiency in traditional methods. The major improvements of the K-medoids algorithm include the following three aspects: (1) Replacing random initial median selection with density-based initialization to reduce the sensitivity to outliers; (2) Integrating a new cluster validity metric that combines intra-cluster compactness and inter-cluster separation to dynamically evaluate the clustering quality during the iterative process; (3) Embedding a LNS to overcome local optimality by iteratively destroying and reconstructing suboptimal clusters. Compared with the genetic algorithm, the improved K-medoids reduced the selection cost by 15.9% and the distribution cost by 13.6%. The time penalty and freshness cost were reduced by 10.4% and 3.0%, respectively. The BWP value of the improved K-medoids model was significantly reduced compared to that of the ant colony optimization. The sensitivity analysis showed that the algorithm was robust under different order sizes and delivery windows. This indicates that the new algorithm provides a scalable solution for dynamic e-commerce logistics by minimizing fulfillment cost while ensuring freshness and timeliness.