Solving diabetes diagnosis problems using machine learning

Abstract

This research is devoted to the study of the use of machine learning methods to solve the problem of diagnosing diabetes. The results of using machine learning in the context of diabetes are varied and depend on the methods of data analysis, the models used and the quality of the data provided. Experiments on the Diabetes dataset were conducted in the study using a Naive Bayes classifier model and a linear kernel SVM for a binary classification problem. Models are trained on the training dataset, standardizing features, and evaluated on the test set using confusion, precision, recall, F1-measure, and AUC-ROC metrics. The results obtained confirm that machine learning can improve the accuracy of diagnosing diabetes and classifying its type. This allows for customized treatment plans to be developed, considering the unique characteristics of each patient. Machine learning models are also successful in predicting the likelihood of complications, allowing for preventative measures to be taken. Their use facilitates the integration of data from various sources, enriching patient information. In conclusion, machine learning-based decision support systems assist physicians and patients in making informed decisions.