Proactive Risk Modeling in Healthcare

Abstract

Proactive risk modeling in healthcare provides a valuable approach for identifying and mitigating potential risks, ultimately improving patient outcomes and enhancing organizational efficiency. Data-driven techniques, such as Bayesian Belief Networks (BBNs), offer a robust framework for modeling uncertainties and capturing interdependencies among clinical variables. In this study, BBNs were used to analyze data from a healthcare survey to determine the relationships between participants' dietary habits, exercise frequency, age, gender, and obesity. The results indicated that calorie consumption monitoring (SCC) and physical activity frequency (FAF) were the variables most influenced by obesity.

Three BBN algorithms were evaluated for their effectiveness in modeling these relationships: Naïve Bayes (NB), Tree-Augmented Naïve Bayes (TAN), and Augmented Naïve Bayes (ANB). Survey data from participants in Mexico, Peru, and Colombia were used to compare these algorithms across several performance metrics, including accuracy, confusion matrix, and F1 score. The findings showed that ANB outperformed the other algorithms, achieving an accuracy and F1 score of 82%. It effectively modeled the complex relationships among the variables, followed by TAN and NB.

Date of Award	9 Dec 2024
Original language	American English
Supervisor	Mecit Simsekler (Supervisor)