Optimized Feature Selection based approach for Predicting Cardiovascular Disease Risk in Indian Population

Cardiovascular disease (CVD) has become a significant concern, being a leading cause of illness and death in many parts of the world. India, being a developing nation, plays a significant role in the burden of cardiovascular disease due to its unique combination of factors such as demography, lifestyle, pollution, genetic tendency, and other influences. Traditional cardiovascular risk calculators may not always provide precise estimates for the Indian population due to variations in risk factors. Hence, there arose a necessity to incorporate machine learning algorithms in this domain. This research suggests an effective cardiovascular disease prediction system tailored specifically for the Indian population, utilizing a novel feature selection approach. This new feature selection method (NFS) is the integration of three different existing methods-Mean Gini decrease (MDG), Recursive feature elimination (RFE) and Mutual Information (MI) based selection. The novelty in the method is that the importance score of each feature was obtained from each traditional feature selection method. These scores were then normalized and added. The top 10 features were selected on the basis of total score. After selecting these 10 features Statistical technique of Pearson Correlation coefficient was used to check the correlation between these features which resulted in selection of optimal feature set of 8 features. Three different machine learning techniques were used to compare the performance of this NFS namely Random Forest (RF), Naïve Bayes (NB) and Logistic Regression (LR). Out of all, RF was found to be the best algorithm with accuracy of 97.5% and sensitivity of 98.2%. The preprocessing techniques used were missing value imputation by Proximity Matrix and Normalization by Min/Max method. K fold cross validation was used with value of K as 10 to avoid any overfitting.