An Ensemble Model for Predicting Cardiovascular Disease utilizing Nature Inspired Optimization

Abstract

This paper represents an efficient model for heart disease prediction model utilizing an ensemble mechanism optimized through BAT algorithm. Worldwide mortality rates are widely acknowledged to be significantly influenced by the prevalence of cardiovascular disease, particularly in economically disadvantaged regions. The need to mitigate the potentially severe repercussions associated with this particular health concern highlights the requirement for accurate and timely outcome prediction. Proposed methodology incorporates Mutual Information for feature selection, Inter Quartile Range for outlier removal. The StandardScaler method is used to achieve feature-wise standardisation in order to mitigate any bias resulting from varying scale disparities. Gradient boosting is an ensemble technique used in model construction that is well-known for its capacity to handle missing data and produce precise predictions. The BAT algorithm is implemented, which further improves speed by utilising optimisation inspired by nature. The application of the BAT method in this particular model has yielded a notable improvement in performance, resulting in an accuracy rate of 84.94%. The precision, specificity, and sensitivity scores of the model were 76.47%, 81.88%, and 89.65%, respectively. These metrics collectively suggest a balanced performance.

Keywords

• BAT algorithm, Cardiovascular disease, Ensemble classifier, Gradient Boosting, Inter Quartile Range, Nature Inspired Optimization