Can We Classify the Participants of a Longitudinal Epidemiological Study from Their Previous Evolution?

Abstract

Medical research can greatly benefit from advances in data mining. We propose a mining approach for cohort analysis in a longitudinal population-based epidemiological study, and show that modeling and exploiting the evolution of cohort participants over time improves classification quality towards an outcome (a disease). Our mining workflow encompasses steps for tracing the evolution of the cohort participants and for using evolution features in classification. We show that our approach separates better between classes and that change in the values of variables is predictive. We report on results for the liver disorder hepatic steatosis (high fat accumulation in the liver), but our approach is appropriate for classification of longitudinal epidemiological data on further disorders.

Monotonicity Detection and Enforcement in Longitudinal Classification

“…Niemann et al [13] proposed an approach for longitudinal classification that used a pre-processing step to cluster all data instances based on their attribute values and generated new predictor attributes based on cluster data. After the clustering is completed and cluster data is added to every instance in each wave, the dataset is transformed by combining all waves into a single dataset, similarly to the two previous studies, omitting the time indexes.…”

3D Regression Heat Map Analysis of Population Study Data

“…They extract classification rules that serve as basis for our proof-of-concept tests. Niemann et al [30] improved the classification performance by generating features (called evolution features) that describe latent temporal information across the study waves. We try to reproduce their results and further investigate findings presented by Niemann et al [31].…”

Constructed Temporal Features for Longitudinal Classification of Human Ageing Data

“…As related work, Niemann et al [3] generated evolution features by comparing instance clustering results at different time points in a longitudinal dataset. In addition, Buizza et al [4] created longitudinal pattern features by comparing distances and means related to two subsequent images (PET/CT scans).…”