Dependent Bayesian Analysis: Application to Text Analysis

Abstract

In this case report, we present a statistical algorithm for predicting influenza using text of reasons for patients' appointment. We introduce two variants of the algorithm: the naïve or Independent Bayesian System, which is based on the assumption of independence among predictors, and the Dependent Bayesian System variant, which assumes dependence among predictors in the data. In Dependent Bayesian System, the sequence of examining the predictors is key, yet little guidance is available in the literature regarding different approaches to selecting the sequence of predictors. In this case report, we describe a maximum likelihood approach to determining such sequence. The resulting procedure is a data-mining algorithm, which is easy to implement in any classification problem with categorical predictors.

locked icon

Sign in to access this content

Get a 30 day FREE TRIAL

  • Watch videos from a variety of sources bringing classroom topics to life
  • Read modern, diverse business cases
  • Explore hundreds of books and reference titles