Analysis of existing options for the classification of patients with cardiovascular disease using neural networks
Keywords:хронобіологія, інформаційна технологія моніторингу, нейронні мережі, класифікація
The work is devoted to the analysis of information technologies of chronobiological monitoring of cardiac systems, development of decision support system for physician-researcher based on classification methods using neural networks such as PNN (Probabilistic Neural Networks), multilayer perceptron and CasCor (Cascade Correlation).
The training sample was 80% of the total number of patients (278 training pairs), and the test sample was 20% (65 training pairs). The presented data sample is highly representative, considering the number of patients: dataset includes 343 patients with cardiovascular complications (class 1 - 142, class 2 - 207). The age range of patients in the sample is 50-60 years old.
The result is the following: the best classifier is the neural network of cascade correlation with 85-88% classification accuracy. The worst classifier was the probabilistic neural network, since the accuracy of this algorithm depends on the size of the data set.
Dziak G.V. Daily blood pressure monitoring / G.V. Dziak, Т.V. Kolesnik, Y.N. Pogoretskii. – Dnipropetrovsk, 2005. – 200 p.
H. Khildebrandt , Chronobiology and Chronomedicine / H. Khildebrandt, М. Mozer, М. Lekhofer. М.: Arnebia. 2006. – 144p.
Bulanaya Т.М. Diagnosis of the cardiovascular system based on probabilistic neural networks / Т. М. Bulanaya // Mpzis-2007: international Research Practice Conf., 14 16 november 2007 р.: thesis. Dnipropetrovsk, 2007. p. 31.
Bulanaya Т.М. Information technologies for monitoring and evaluating the efficiency of diagnosing the condition of technical and cardio systems: Dis, … Candidate of Technical Sciences: 05.13.06: Defended 27.06.2018; Approved 23.10.2018. D., 2018. 145 p.
Colombet I., Ruelland A., Chatellier G., Gueyffier F., Degoulet P., Jaulent M.C. Models to predict cardiovascular risk: Comparison of CART, multilayer perceptron and logistic regression. Proceedings AMIA Symposium , 2000. 156–160.