Investigation of the efficiency of graph data representation for a cardiovascular disease predictive model by deep learning methods

Currently, cardiovascular diseases (CVD) are the most common cause of death in the world. Artificial intelligence methods provide extensive opportunities for extracting new knowledge from the raw data of medical information systems (MIS). This study is aimed at building a model for predicting the diagnosis of CVD based on patient complaints at a doctor's appointment using natural language processing methods. The formation of the initial data set is based on a graph model of the patient's medical history with CVD according to the visit protocols. A comparative analysis of machine learning models such as the naive Bayesian classifier, the support vector machine and convolution neural networks is carried out. As a result of the experiments, the most effective model for predicting CVD has been selected.