Research of Algorithms for Binary Classification of Biomedical Data for Recognition of Periodontal Tissue State

The aim of the study is to analyze algorithms for binary classification of biomedical data for recognizing the state of periodontal tissues based on the obtained electrochemical parameters. To collect data, an information-measuring system developed by the authors was used to assess the electrochemical characteristics of periodontal tissues. As a result of the pilot study conducted by the authors, a database was created. The database contains 214 records and consists of four fields: age, gender, current work, group. The pilot study involved 214 patients aged 19 to 25 years, including 123 men and 91 women. The development and analysis of binary classification algorithms was carried out in the Microsoft Machine Learning Studio (classic) environment. The model consists of three blocks: data preparation, training of binary classifiers, evaluation of the results obtained and includes the following algorithms: Two-Class Averaged Perceptron, Two-Class Bayes Point Machine, Two-Class Boosted Decision Tree, Two-Class Decision Forest, Two -Class Decision Jungle, Two-Class Logistic Regression, Two-Class Neural Network, Two-Class Support Vector Machine. As a result of the analysis of the algorithms, a table was compiled containing the main evaluation metrics: Accuracy, Precision, Recall, F1 Score, Area Under Curve (AUC). The analysis of the table showed that within the framework of the task of binary classification of electrochemical parameters of biological objects, the Two-Class Bayes Point Machine and Two-Class Support Vector Machine are the most preferable.