Smart Forecasting of Blood Pressure Using Machine Learning and Physiological Insights

Abstract

Hypertension has important health consequences that require reliable predictive models. This project applies Python and the Random Forest Classifier to predict blood pressure in patients who complete surveys related to their age, BMI and amount of aerobic exercise. The data preprocessing ensures that the data was legitimate before the model was provided data to learn from. The classifier was 100% correct in the training data and 86% for testing data, suggesting very good predictability. The model appears to be able to maintain sufficient generalization, which represents a valuable tool for the early identification of hypertension and a decision-use of healthcare knowledge. This machine learning approach was able to validate our non-linear relationships obtained from previous knowledge about the clinical data, and demonstrates the potential for data-oriented approaches to transform the landscape of cardiovascular risk management