Determination of Factors Affecting Severity of Helicobacter pylori for Gastric Biopsy Samples by CART Decision Tree Algorithm

Abstract

Objective: H. pylori wich is one of the important gastric pathogens and is a motile, non-sporeless, encapsulated, microaerophilic, gram-negative bacterium. The aim of this study was to determine the factors affecting disease severity in patients with a positive pathologic diagnosis of Helicobacter pylori after gastric biopsy by data mining. It was aimed to utilize the more descriptive structure of data mining algorithms compared to traditional classification and regression approaches. Methods: The study data were obtained from gastric biopsy samples of 1247 patients, 40.5% male and 59.5% female, who were sent to the pathology laboratory between 2014 and 2018. A total of 6 factors including age, gender, inflammation, metaplasia, atrophy and activation, which are thought to have an effect on gastric H. pylori severity, were examined. Querying the effects of factors was done with the CART (Classification and Regression Trees) decision tree algorithm, one of the data mining algorithms. Results: The factors ranking as their effect on the severity of gastric h. pylori, as follows; activation > inflammation > metaplasia > atrophy > age > gender in a percentage of normalized importance at 100.00%, 88.6%, 51.4%, 38.1%, 12.8%, 3.3% respectively. Conclusion: As a result, levels of activation, inflammation, and metaplasia emerged as the most important factors affecting gastric H. pylori severity.