PREDICTION OF BLOOD PRESSURE LEVEL USING NEURAL NETWORK MODELS

Psychophysiological changes are common factors contributing to the development of arterial hypertension. The aim of this study was to examine the psychophysiological characteristics of individuals with different levels of blood pressure (BP), to build and compare the predictive accuracy of a classical neural network and a Kolmogorov-Arnold network in forecasting BP levels, as well as to determine the prognostic value of the studied psychophysiological parameters. The study involved 240 practically healthy individuals aged 18 to 22. All participants underwent initial BP measurements, based on which they were divided into three groups according to the recommendations of the European Society of Cardiology. The research included 24-h BP monitoring, an assessment of anxiety levels, and evaluations of well-being, activity, mood, and sleep quality. Predictive models were created from the obtained data to forecast systolic BP levels >130 mmHg. Situational and trait anxiety levels were found to be significantly higher in individuals with normal-high BP compared to those with normal-low and normal BP levels. Sleep quality, measured by the PSQI (Pittsburgh sleep quality index) questionnaire, was significantly lower in individuals with normal-high BP compared to the other groups. The neural network constructed in this study demonstrated that psychophysiological indicators can be effectively used for predicting elevated BP levels and for the early diagnosis of arterial hypertension. This research is the first to apply the Kolmogorov-Arnold network for predicting high BP levels. The study found that this network was highly effective, outperforming the multilayer perceptron with a larger number of neurons in terms of predictive accuracy.