In this article, we focus on addressing the challenges of preprocessing training samples for pharyngitis and hepatitis in the medical field. Initially, we propose algorithms to tackle issues such as normalizing the symbolic values of training sample objects and converting these values into binary form (0 or 1). After this, a standard training sample is formed by classifying objects of the class, selecting informative symptom complexes, assessing the classification efficiency of the selected informative symptom complexes, and forming informative symptom complexes in sets of informative features. Using the software built upon these developed algorithms, practical medical problems were solved in a comprehensive manner. Finally, as a result of solving these problems, that is, the problem of evaluating the classification effectiveness of selected informative symptom complexes for pharyngitis and hepatitis diseases, the results were obtained and comparatively analyzed using existing algorithms on the famous KNIME analytical platform.