GWO-XGB: Grey Wolf Optimization-based eXtreme Gradient Boosting for Hypertension Prediction in Bangladesh

Author:- Tasfia Tahsin, Khondoker Mirazul Mumenin, Farhana Tazmim Pinki, Anamika Biswas Tuli, Shahriar Sikder, Md. Ashfikur Rahman, Abdullah Al-Mamun Bulbul, Md. Abdul Awal
Category:- Conference; Year:- 2021
Discipline:- Development Studies Discipline
School:- Social Science School

Abstract

Hypertension is rapidly increasing day by day worldwide as well as in Bangladesh. The majority of people in our country die due to hypertension. So, early prediction of this disease is a very important task that may reduce the number of affected patients. In our paper, we have proposed the Grey Wolf Optimization-based eXtreme Gradient Boosting (GWO-XGB) model which can predict hypertension based on Bangladeshi data collected in 2017–18 (BDHS'17-18). Here the hyperparameters of XGB are optimized using GWO and we have attained 91.26% accuracy in hypertension classification using this model. We have calculated performance evaluation metrics (Accuracy, error score, F1 score, Kappa score, MCC score, sensitivity, specificity) and plotted the precision-recall curve, bootstrap ROC curve to compare the performance of the proposed GWO-XGB model with some of the state-of-the-art classifiers. This study has also calculated the most influencing features to predict hypertension which will assist national policy-maker to provide more emphasis to those significant features.

Read More