Study of anomaly registration detection based on multilayer kernel autoencoder extreme learning machine model.

Area of Science:

• Health Informatics
• Artificial Intelligence
• Machine Learning

Background:

• Digital transformation in hospitals is driven by AI and health informatics.
• Improper registration practices lead to uneven allocation of medical appointment resources.

Purpose of the Study:

• To propose an integrated machine learning algorithm (DKELM-PSS) for detecting anomalous registration behavior in hospital health information systems (HIS).
• To enhance the allocation of medical resources and promote intelligent hospital development.

Main Methods:

• Data preprocessing using Sparse Principal Component Analysis (Sparse PCA) for denoising and dimensionality reduction.
• Deep feature extraction using a Deep Kernel Extreme Learning Machine (DKELM) model with stacked Kernel Extreme Learning Machine Autoencoders (KELM-AE).
• Optimal parameter configuration using the Salp Swarm Algorithm (SSA) for improved classification accuracy and stability.

Main Results:

• DKELM-PSS achieved an optimal accuracy of 0.9942 in HIS anomaly detection.
• Experimental results validated the effectiveness and robustness of DKELM-PSS compared to SVM-RBF, XGBoost, and ResNet.
• The method demonstrates superior performance in identifying anomalous registration behavior.

Conclusions:

• The proposed DKELM-PSS offers an efficient anomaly detection method for hospital HIS data.
• This approach is conducive to optimizing medical resource allocation.
• The study contributes to the intelligent development of hospital services through advanced AI techniques.