Optimization of hospital resource scheduling efficiency based on dynamic weighted distance anomaly detection algorithm.

This study developed an intelligent resource scheduling integrated model to address engineering challenges like load imbalance, low energy efficiency, and delayed anomaly detection in hospital communication and energy-saving systems. This model integrated the Dynamic Weighted Distance-Based Outlier Detection (DWD-OD) algorithm based on dynamic weighted distance with the Multimodal Resource Collaborative Optimization (MRCOF) framework at the system level. The DWD-OD algorithm improved the distance measurement method through engineering means, and constructed a dynamic weight matrix by introducing device types and load characteristics to adapt to the scale differences of heterogeneous data. By combining sliding time windows with local density threshold mechanisms, this scheme achieved real-time recognition of energy consumption events such as heating, ventilation, and air conditioning (HVAC) overload or lighting anomalies. Experimental data showed that the detection accuracy of this method reached 95.3%. To achieve resource optimization, the model integrated an improved Genetic Simulated Annealing Algorithm (GSAA), forming a two-stage optimization strategy: Stage 1 managed virtual resources through OpenStack to isolate abnormal tasks; Stage 2 utilized Kubernetes for path optimization of containerized infrastructure to reduce communication latency and mitigate fragmentation risks. Based on a real dataset validation of the communication and energy system in a tertiary hospital, it was shown that under the hybrid cluster architecture, the overall energy efficiency of the system was improved to 89.7%, and the response time for critical tasks was controlled within 4.8 s. The integrated Development, Security, and Operations (DevSecOps) lightweight security control layer had increased vulnerability scanning efficiency by 27.3%. It should be pointed out that the validation of this study was only based on data from a single medical center, and some abnormal samples were enhanced by generative adversarial networks. Security assessment depends on the model simulation environment, and these factors may have certain applicability biases when extended to medical institutions of different sizes.