IoT and Multi-Source Data-Driven Intelligent Warehouse Optimization Using KMA-RDTC

Abstract

Warehouse operations have developed over the decades through the enhanced Internet of Things (IoT) technology. The traditional operation is perceived as inconsistent and operationally inefficient, which are the main deficiencies in traditional inventory processes. The research fills the gaps by introducing a smart warehouse optimization tool on IoT and multi-source data to optimize and refine inventory accuracy, reduce operational expenses, and optimize the use of resources. Data from 2500 real-time warehouse operations were pre-processed using min-max normalization, and dimensionality reduction was applied via Principal Component Analysis (PCA). The Komodo Mlipir Algorithm-tuned Random Decision Tree Classifier (KMA-RDTC) was proposed to integrate the RDTC with the KMA for optimizing warehouse operations. The RDTC provides robust classification and anomaly detection for inventory data and demand forecasting, while the KMA optimizes RDTC hyperparameters using exploration-exploitation strategies, enabling efficient real-time anomaly detection and forecasting, and enhancing predictive performance, operational accuracy, resource utilization, and decision-making in intelligent warehouse environments. Through experimental testing implemented in Python, it demonstrated superior predictive reliability over traditional Machine Learning (ML) methods by providing accuracy (0.985), prediction accuracy (0.958), sensitivity (0.934), specificity (0.962), Root Mean Squared Error (RMSE) (0.10), and Mean Absolute Error (MAE) (0.07) with Python implementation. These results indicate accurate anomaly detection, efficient resource utilization, and high operational precision. The framework provides a comprehensive design for intelligent warehouse operations, enabling scalable, data-driven, and sustainable logistics aligned with Industry 4.0 principles.