A Machine Learning Prediction Model Based On Heterogeneous Temporal Data In Electronic Health Records

Health monitoring sensor-cloud system(SCS)integrates sensors,sensor networks and cloud,which is used to manage sensors,collect and process human sign parameter data,and make decisions based on the processed data,so as to realize the monitoring of human health status and the prevention and diagnosis of diseases.Though the SCS has received tremendous attention from both academia and industry because of its numerous exciting applications,it still faces the challenge in reliability.In this work we make novel contributions by proposing a network simplification method based on graph decomposition and reconstruction through articulation vertices,which effectively remove all redundant network edges and vertices,leading to a significantly reduced network model for accurate and efficient K-terminal reliability analysis.Empirical studies show that the proposed simplification method integrated with the binary-decision-diagrams based evaluation algorithm can significantly speed up K-terminal reliability analysis of large real-life SCS.Electronic health records(EHR)contain a large number of longitudinal data,which is valuable for biomedical informatics research.However,standard learning algorithms present challenges due to the complex structure of the data.Some methods of temporal data modeling lead to the loss of potentially valuable sequential information.In this paper,a new representation of temporal data in EHR are studied,which preserves the sequential information that can be processed directly by the standard machine learning algorithms.Empirical studies using clinically measured datasets in the real-life database of EHR have shown that using distance metrics for random subsequences significantly improve predictive performance compared to the use of original sequences or clustering sequences.Predictive models built using temporal data in EHR can potentially play a major role in improving management of diseases.Due to the sequence correlation and large feature space dimensions,traditional methods such as machine learning and non-deep neural networks are difficult to provide accurate predictions of disease.Recent works show that the long short term memory(LSTM)neural network outperforms most of those traditional methods for disease prediction problems.In this study,a hybrid deep learning neural network framework that combines Inception with the attention mechanism LSTM is proposed to further improve the prediction accuracy.Empirical studies using the real-world datasets in EHR have shown that using the proposed hybrid deep learning neural network for disease prediction significantly improves predictive performance compared to the use of SVM model,CNN and LSTM alone.