Machine learning methods for analyzing medical data

Informace o projektu

Poskytovatel
České vysoké učení technické v Praze
Program
Studentská grantová soutěž ČVUT
Číslo projektu
SGS20/132/OHK4/2T/14
Zahájení projektu
Ukončení projektu
Popis projektu

Bayesian networks are examples of probabilistic graphical model successfully used in a variety of real-world applications where it is necessary to support decision-making under uncertainty. The basic advantage of Bayesian networks is that they allow using a directed graph to model the relationships between variables and then use these relations for efficient computation of conditional probabilities in the model (i.e., for probabilistic inference). This allows the use of Bayesian networks in applications where it is necessary to model relationships among hundreds of variables. Project "Machine learning methods for analyzing health data" will be based on real data on patients with acute myocardial infarction. The first step was to test known statistical models for classification, regression, and probabilistic modeling to predict treatment results and model the cost of the treatment and we have succeeded in building a modification of the TAN algorithm which deals with incomplete and imbalanced data. The aim of this phase is to learn the optimal BN structures from incomplete and imbalanced data.