MACHINE AND DEEP LEARNING METHODS IN APPLIED TASKS OF BIOLOGY AND MEDICINE
Igor S. Golyak, 1
Igor L. Fufurin, 1
Dmitriy R. Anfimov, 1,
Pavel V. Berezhanskiy, 2,
Andrey Morozov, 1
1 Bauman Moscow State Technical University, 105005, Moscow, Russia
2 Morozov Children’s Clinical Hospital, State Budgetary Healthcare Institution, Moscow Healthcare Pulmonology Department, Moscow 119049, Russia
Abstract
The report discusses the use of machine and deep learning methods for the early diagnosis of chronic diseases by the volunteers exhaled air. In particular, the possibilities of machine and deep learning methods are being explored in the classification of the following chronic diseases: type 1 diabetes, pneumonia, and asthma. Comparative analysis includes the use of dimensionality reduction methods (PCA, t-SNE) with classification methods (support vector machine, logistic regression, random forest) from machine learning and convolutional neural networks and autoencoders from deep learning methods. It is shown that the combination of autoencoders with SVM methods and logistic regression makes it possible to achieve a classification accuracy close to 100%.
Speaker
Igor Golyak
Bauman Moscow State Technical University
Russia
Discussion
Ask question
