A Review of the Utility of Bayesian Network Models

Bayesian Networks are probabilistic models built from conditional probability tables that relate two observable instances to one another in parent-child fashion. The networks’ strength lies in their ability to use inferential logic to make likelihood assessments about a parent node based on an observation of its child. Additionally, they make it very easy to combine quantitative data with qualitative knowledge from industry experts. These abilities make them very attractive for use as formulation tools in the paint and rubber industries. Paint and rubber formulation has long proven to be a challenging task because companies have a difficult time compiling the data from all their formulators- data that often contains large amounts of opinion. This paper seeks to define Bayesian Networks and a few inferential operations using them, and then to apply these methods to three distinct industry problems. This paper explores applications including: (1) marketing, (2) expert knowledge collection, and (3) a traditional formulation study. This paper is submitted as part of graduation requirements for the University of Akron Williams Honors College, 2018