Bayesian Classification of Self-Reported Gastrointestinal Conditions among Mathematics Students

Abstract

This study examines a simple Bayesian classification procedure using questionnaire summaries from mathematics students at Universitas Sumatera Utara. Four self-reported categories were considered: dyspepsia, gastroesophageal reflux disease, gastritis, and Helicobacter pylori infection. The original charts contain category totals of 13, 24, 3, and 1, respectively, representing 41 questionnaire classifications. Fifteen symptom indicators were tabulated, and a worked example combined six reported symptoms under the naive conditional independence assumption. Recalculation using the manuscript’s rounded inputs yielded normalized model probabilities of 4.52% for dyspepsia and 95.48% for gastroesophageal reflux disease, with zero weights for the other categories. These zeros resulted from unsmoothed zero-frequency estimates. A supplementary calculation using chart-derived counts produced 95.38% for gastroesophageal reflux disease; adding one to each binary symptom count reduced this value to 80.18%. The comparison shows that numerical confidence depends on estimation choices in sparse categories. The results demonstrate a reproducible probability calculation, but do not establish diagnostic accuracy. Self-reported labels, overlapping conditions and symptoms, incomplete recruitment information, and the absence of independent clinical validation limit interpretation. The procedure is therefore presented as an educational classification example requiring further data and validation before any clinical application.