In recent years, heart disease has become a reason for high mortality rate, and data mining has also gained attention in the medical domain. Predicting this disease in its initial stage helps to save lives and reduce treatment costs. Various classification models were recently introduced with expected outcomes. However, they lacked prediction accuracy. Hence, the aim of this study was to employ data mining techniques for predicting heart disease, and focused on higher accuracy. This disease was predicted by considering the Cleveland heart disease dataset, employing deep CNN models for extracting relevant features, and performing feature level fusion related to its efficient and automatic learning. FGM-PCA (Fast Track Gram Matrix-Principal Component Analysis) was proposed for dimensionality reduction and fusion to solve overfitting issues, minimise time and space complexity, eliminate redundant data, and enhance classifier performance. Further, effective classification was achieved through the newly introduced IEB-RF (Informative Entropy Based-Random Forest) because it offers high accuracy and can also handle a large amount of data flexibly. The proposed system was evaluated in terms of accuracy, sensitivity, F1-score, AUC (Area under Curve) and precision. The results revealed the superior performance of the introduced system in comparison to traditional techniques.
Keywords: heart disease prediction; FTGM-PCA; informative entropy based-random forest; dimensionality eduction; cleveland heart disease dataset
*Corresponding author: Tel.: (+91) 9647533289
E-mail: deepikaphd11@gmail.com
