Enhancing Clustering Performance in Sepsis Time Series Data Using Gravity Field.

Sepsis, a severe systemic response to infection, represents a pressing global public health challenge. Time series research, including the analysis of medical data, encounters significant obstacles due to the high dimensionality, complexity, and heterogeneity inherent in the data associated with sepsis. To address these obstacles, this paper proposes a novel approach for enhancing time series datasets. The primary objective of this approach is to enhance clustering performance and robustness without requiring modifications to existing clustering techniques. Specifically, this approach can improve the clustering performance in sepsis patients. The effectiveness of the proposed approach is validated through comprehensive experiments conducted on both non-medical and medical sepsis datasets, showcasing its potential to advance time series analysis and significantly contribute to the effective management of sepsis medical conditions. In addition, we use this approach to establish three subtypes in the clustering of sepsis patients, which provide meaningful interpretations in terms of medical significance and we further explore the therapeutic heterogeneity among the three subtypes.