Prediction of BP reactivity to talking using hybrid soft computing approaches

Area of Science:

• Biomedical Engineering
• Cardiovascular Research
• Data Science in Medicine

Background:

• High blood pressure (BP) increases cardiovascular disease risk.
• Precise BP measurement is vital for clinical and research settings.
• Anthropometric characteristics can influence BP reactivity.

Purpose of the Study:

• To predict blood pressure (BP) reactivity to talking using anthropometric data.
• To evaluate the efficacy of different predictive models.
• To address multicollinearity among predictor variables.

Main Methods:

• Utilized anthropometric data (age, height, weight, BMI, AC) as predictors.
• Applied Principal Component Analysis (PCA) to manage multicollinearity.
• Integrated PCA with Artificial Neural Network (ANN), ANFIS, and LS-SVM models.
• Compared model performance using R², RMSE, and MAPE.

Main Results:

• The PCA-LS-SVM model demonstrated superior prediction accuracy for BP reactivity.
• PCA effectively reduced multicollinearity among anthropometric variables.
• PCA-fused models outperformed individual models in predicting BP changes.

Conclusions:

• PCA-fused prediction models offer significant advantages for biological variable prediction.
• The PCA-LS-SVM model provides a robust method for forecasting BP reactivity.
• Accurate BP reactivity prediction aids in cardiovascular risk assessment.