| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 2.86 MB | Adobe PDF |
Authors
Abstract(s)
The drive to understand the laws that govern the universe and ourselves in order to expand our view of reality is deeply rooted in humanity. In science, this urge is a robust process filled with challenges and opportunities given the rapidly growing technology-driven volume of time series data. Causal discovery supports science in an innovative and fast-growing manner with the essential goal of uncovering mathematical orders directly from
observational data translated into causal association networks. This scientific tool pledges to accelerate growth in various fields, including life sciences. This work approaches the topic of causal discovery on two levels. First, we address the theory of constraint-based methods on detecting and quantifying causal relations, covering how the methods work, the challenges they face, and the opportunities they present. Second, we explore the PCMCI method with an implementation on both synthetic and real-world data. The results of this work found in applying causal discovery in real physiological signals data may provide insights into the prospects and difficulties of causal structure search in healthcare Big Data and, moreover, the advantages of using causal models in prediction.
Description
Keywords
Causal discovery Causal relations Time-series data Physiological signals MIMIC III