Библиографическое описание:Cancer Prediction Models Using Gene Expression and Logical Analysis of Data : доклад, тезисы доклада / M. Bartosh, I. Masich. - [S. l. : s. n.], 2022. - Текст : непосредственный // HYBRID METHODS OF MODELING AND OPTIMIZATION IN COMPLEX SYSTEMS : Proceedings of the International Workshop “Hybrid methods of modeling and optimization in complex systems” (HMMOCS 2022) / Krasnoyarsk Regional Science and Technology City Hall, Russia Siberian Federal University, Reshetnev Siberian State University of Science and Technology ; International Workshop “Hybrid methods of modeling and optimization in complex systems” (HMMOCS 2022) (2022 ; 22.11 - 24.11 ; Krasnoyarsk). - London, United Kingdom : European Proceedings, 2022. - P. 379-386. - ISBN 9781802969603, DOI 10.15405/epct.23021.47.
Аннотация:The paper analyzes gene expression data in relation to the diagnosis and prognosis of the development of oncological diseases. The goal is to create a hybrid prediction model based on gene expression data and interpretive machine learning. The experiments were carried out on four publicly available gene expression datasets in relation to the prediction of breast and lung cancers. Data sets contain information about positive and negative observations, described by tens of thousands of attributes with gene expression data. Logical analysis of data is investigated as the main method for building a model. This method is based on combinatorics and optimization. As a result of logical analysis of data, a set of patterns is built, each of which involves only a small number of input attributes (genes). The search for a reference set of attributes, which is a step in the logical analysis of data, yields a small number of genes that have a combinatorial effect on the result. The resulting patterns have a small number of conditions and are understandable to the user. A comparison was made with other machine learning algorithms, including rule based classifiers: RIPPER, decision trees, and others. Logical analysis of data has advantages both in terms of classification accuracy and result interpretability, and therefore provides greater confidence in the recognition result.
