DBN-Extended: A Dynamic Bayesian Network Model Extended With Temporal Abstractions for Coronary Heart Disease Prognosis

Abstract

Dynamic Bayesian networks (DBNs) are temporal probabilistic graphical models that model temporal events and their causal and temporal dependencies. Temporal abstraction (TA) is a knowledge-based process that abstracts raw temporal data into higher level interval-based concepts. In this paper, we present an extended DBN model that integrates TA methods with DBNs applied for prognosis of the risk for coronary heart disease. More specifically, we demonstrate the derivation of TAs from data, which are used for building the network structure. We use machine learning algorithms to learn the parameters of the model through data. We apply the extended model to a longitudinal medical dataset and compare its performance to the performance of a DBN implemented without TAs. The results we obtain demonstrate the predictive accuracy of our model and the effectiveness of our proposed approach.