Probabilistic change detection and visualization methods for the assessment of temporal stability in biomedical data quality

摘要: Knowledge discovery on biomedical data can be based on-line, data-stream analyses, or using retrospective, timestamped, off-line datasets. In both cases, changes in the processes that generate their quality features through time may hinder either knowledge process generalization of past knowledge. These problems seen as a lack temporal stability. This work establishes stability dimension and proposes new methods for its assessment probabilistic framework. Concretely, are proposed (1) monitoring changes, (2) characterizing trends detecting subgroups. First, change detection algorithm is Statistical Process Control posterior Beta distribution Jensen---Shannon distance, with memoryless forgetting mechanism. (PDF-SPC) classifies degree current three states: In-Control, Warning, Out-of-Control. Second, novel method to visualize characterize projection non-parametric information-geometric statistical manifold windows. facilitates exploration IGT-plot and, by means unsupervised learning methods, discovering conceptually-related Methods evaluated real simulated National Hospital Discharge Survey (NHDS) dataset.