From Complexity to Clarity: Clustering and Classification for multidimensional functional structure
開催期間
14:00 ~ 15:00
場所
講演者
概要
The growing prevalence of multi-dimensional and longitudinal data in clinical research presents both challenges and opportunities. Addressing the inherent complexity of such data requires advanced methodologies that can reveal meaningful patterns and insights. This presentation introduces innovative clustering and classification approaches designed specifically for functional data analysis, bridging the gap between complexity and clarity.
We explore advanced clustering techniques, including Multi-dimensional Fréchet K-means for Longitudinal Data (MFKmL) and Sparse Fréchet K-medoids (SFKmL), which adeptly handle irregular and asynchronous trajectories. These methods segment patient data into clinically relevant groups based on trajectory shapes and variable contributions, enabling personalized insights.
Additionally, the Sparse Functional Linear Discriminant Analysis (SFLDA) framework for multivariate gait data classification is highlighted. By leveraging the functional nature of gait cycles and employing regularization techniques, SFLDA not only classifies but also interprets complex gait patterns in patients with cerebral palsy.
Using real-world case studies—thyroid cancer clustering and gait pattern classification—the presentation demonstrates how these techniques uncover actionable insights for precision medicine. Attendees will gain a deeper understanding of the power of functional data analysis to transform the complexity of clinical datasets into clear, interpretable outcomes.