Comparative Analysis of LASSO and RFE for ML-Based Hypertension Risk Prediction Framework

Hypertension constitutes a major worldwide health concern that can result in cardiovascular disease, cerebrovascular accidents, kidney failure, and early death. Hypertension represents a significant global health issue that frequently occurs asymptomatically. This increases the likelihood of patients experiencing significant cardiovascular problems. Anticipatory forecasting of events is essential for prompt reaction and effective management. This study presents a hybrid framework designed for predicting hypertension, utilising the High-Fidelity Synthetic Dataset for Hypertension Prediction. This work implements and compares two feature selection methods Least Absolute Shrinkage and Selection Operator (LASSO) and Recursive Feature Elimination (RFE), for optimized feature selection which leads to better forecast of hypertension. The selected features are then fed into five different machine learning models XGBoost, SVM, Random Forest, MLP and CatBoost for training and testing and analysis purpose. An extensive analysis is done on both the frameworks LASOO + ML models and RFE + ML models including exploratory data analysis, computation of various standard matrices, calibration analysis and explainable AI analysis. All these results indicate that LASOO + XGB model outperform every other integrated model by achieving 98.56% accuracy. The best performance model LASOO + XGB is also compared with the other similar work, where it also outperforms. From all the analysis done we can conclude that the proposed model aims to enhance and helps the healthcare providers in hypertension risk prediction.