PROGRAM APPLICATION FOR STRESS LEVEL MONITORING, BASED ON CLASSIFICATION MODELS
Keywords:stress monitoring, electrocardiography, machine learning, heart rate variability, classification algorithms, bio signals, decision tree, random forest
The paper considers studies, regarding the stress level. After COVID-19 pandemia and constant stay in the forced isolation, the level of stress became higher due to the increase of anxiety. That is why, the study of the basic mechanisms of the stress and monitoring of different biophysiological and bio-chemical reactions of the organism on the stress is of great interest for the researches. Reliable biomarker or stress indicator could provide accurate monitoring of stress, potentially enabling to avoid pathological states at the early stages. Long lasting stress may have negative health outcomes. Thus, the ability to determine when a person is in the state of stress, may be very useful to avoid health problems, especially for patients with suicidal thoughts.
The given study contains the results of the stress level monitoring by means of using the classification models as the forecasting and as a bio signal – heart rate variability (HRV) from the sensors of electrocardiography. Correlation of all variables was carried out , so that only those variables which have high correlation with stress could participate in the models learning. For achieving the set task the following methods have been used: artificial neural network, k-nearest neighbors (KNN), random forest, decision tree. Classification model random forest obtained the highest index of the forecasting accuracy of the presence or absence of the stress – 98 %. On the base of this model program application in programming language R with users interface was developed, it allows to load the data of the electrocardiogram and obtain the conclusion, regarding the level of the stress level. By means of the application the user can control the level of personal stress and lead a healthy life.