Constructing Interpretable Classifiers to Diagnose Gastric Cancer Based on Breath Tests

Quick, inexpensive and accurate diagnosis of gastric cancer is a necessity, but at this moment the available methods do not hold up. One of the most promising possibilities is breath test analysis, which is quick, relatively inexpensive and comfortable to the person tested. However, this method has not yet been well explored. Therefore in this article the authors propose using transparent classification models to explain diagnostic patterns and knowledge, which is acquired in the process. The models are induced using decision tree classification algorithms and RIPPER algorithm for decision rule induction. The accuracy of these models is compared to neural network accuracy.