Topic modeling for systematic review of visual analytics in incomplete longitudinal behavioral trial data

Smart Health (Amst). 2020 Nov:18:100142. doi: 10.1016/j.smhl.2020.100142. Epub 2020 Nov 13.

Abstract

Longitudinal observational and randomized controlled trials (RCT) are widely applied in biomedical behavioral studies and increasingly implemented in smart health systems. These trials frequently produce data that are high-dimensional, correlated, and contain missing values, posing significant analytic challenges. Notably, visual analytics are underdeveloped in this area. In this paper, we developed a longitudinal topic model to implement the systematic review of visual analytic methods presented at the IEEE VIS conference over its 28 year history, in comparison with MIFuzzy, an integrated and comprehensive soft computing tool for behavioral trajectory pattern recognition, validation, and visualization of incomplete longitudinal data. The findings of our longitudinal topic modeling highlight the trend patterns of visual analytics development in longitudinal behavioral trials and underscore the gigantic gap of existing robust visual analytic methods and actual working algorithms for longitudinal behavioral trial data. Future research areas for visual analytics in behavioral trial studies and smart health systems are discussed.

Keywords: MIFuzzy; Missing data; Systematic review; Topic modeling; Visual analytics; Visualization.