Helicobacter pylori Detection Using Machine Learning Algorithm

Abstract

It is well known that Helicobacter pylori (H. Pylori) is a major cause of chronic active gastritis in both children and adults. There are a variety of tests for detection of H. pylori infection, however, in medicine, the only way to diagnose the existence of H. pylori microbe is doing endoscopy which is painful and insufferable for young children [1]. To solve this problem, some machine learning classifiers have been used here to diagnose the existence of this infection. As we will see, using machine learning classifier for diagnose the existence of H. pylori is an alternative method to avoid painful endoscopy. One hundred patient related data has been used from previous published study. There are twenty features in this dataset, such as: abdominal pain and nausea. We have further investigated the contribution of each single feature by using leave-one-feature-out model, where in each experiment one feature was removed from all features model. This model can help us to see how the features interact and how the most and the least informative features can be found, respectively.