Real-Time Order Verification in High-Velocity Micro-Fulfillment Centers using Probabilistic Weight Aggregation and Gaussian Mixture. Models

The operational architecture of Quick Commerce (Q-Commerce) demands order fulfillment within strictly compressed time windows (60–120 seconds), often compromising verification accuracy. Traditional deterministic weight validation methods, which rely on static "Master Data" summations, suffer from high failure rates due to the stochastic nature of inventory—specifically, the high variance in fresh produce and unannounced packaging drifts. This paper proposes a computational framework for Probabilistic Weight Aggregation using Gaussian Mixture Models (GMM). By modeling Stock Keeping Units (SKUs) as multi-modal probability distributions rather than scalar constants, the proposed system dynamically adapts acceptance thresholds based on basket composition. Using a Python-based simulation on a synthetic dataset of 50,000 orders—engineered to mirror the inventory complexity of Indian Q-Commerce platforms (e.g., Swiggy Instamart, Blinkit)—the algorithm demonstrates a 14.2% reduction in False Rejection Rate (FRR) and a 9% improvement in Anomaly Detection (Recall) compared to static thresholding . The system achieves an inference latency of