An Application of Bayesian Network in Cognitive Behavioral Therapy

Abstract

Mental health has received increased focus in recent years, with a larger emphasis on treatment and acceptance. However, evidence-based psychological interventions are of poor availability and have room for improvement. The amount of data being gathered across applications and practices provide opportunities for deeper analysis through machine learning based technologies. By applying Bayesian networks (BNs) in a cognitive behavioral therapy for adults with ADHD, this research analyzes historic self-report data to predict the behavior of future participants at an early stage of the online intervention. Bayesian networks represent probabilistic models that describe the joint probability distribution through an acyclic graph. The contribution of this thesis is an artifact with the purpose of serving as a decision making support tool. Methods of Design Science Research was applied to achieve this, in a development cycle with three main iterations.

Using Bayesian networks for analyzing behavioral patterns yield positive results with its predictive capabilities when dealing with uncertainty. Domain experts from the internet-delivered intervention provided useful feedback and insight that contributed to the novelty and research scope of this thesis. Future work should update the model when a larger population sample is available, and focus on implementing the artifact in a more user-centered desktop application.