Decision Support based on Bio-PEPA Modeling and Decision Tree Induction

Abstract

The problem of selecting determinant features generating appropriate model structure is a challenge in epidemiological modelling. Disease spread is highly complex, and experts develop their understanding of its dynamic over years. There is an increasing variety and volume of epidemiological data which adds to the potential confusion. The authors propose here to make use of that data to better understand disease systems. Decision tree techniques have been extensively used to extract pertinent information and improve decision making. In this paper, the authors propose an innovative structured approach combining decision tree induction with Bio-PEPA computational modelling, and illustrate the approach through application to tuberculosis. By using decision tree induction, the enhanced Bio-PEPA model shows considerable improvement over the initial model with regard to the simulated results matching observed data. The key finding is that the developer expresses a realistic predictive model using relevant features, thus considering this approach as decision support, empowers the epidemiologist in his policy decision making.

Improving process algebra model structure and parameters in infectious disease epidemiology through data mining

“…In contrast, data mining provides individual-level information through analysis of observed data to extract key model features, and analysis of complex and varied time series outputs for a range of parameter settings. The combination of process algebra and data mining is only beginning to be explored [14,15] and plays to the complementary strengths of each.…”