Hybrid prediction model with improved score level fusion for heart disease diagnosis

Abstract

Heart disease diagnosis is a challenging task, which provides an automated forecast of the patient's heart illness to make future treatment simpler. This has led to extensive interest in heart disease diagnostics in the medical sector. However, as there are various risks, the prediction must be more appropriate to avoid death. This work intends to develop the Hybrid Prediction Model with Improved Score Level Fusion (HPISLF) for Heart Disease Prediction. Preprocessing is the first process, where improved min-max normalization is done to preprocess the input data. Feature extraction plays a major role as it extracts additional information from the input data via extracting HOS, Improved Holoentropy-based features, and MI are extracted. Also, proposing a hybrid classification model for diagnosis, which trains the model with the extracted feature set. The final classification outcome is determined by the improved score level fusion that fuses the classification outcomes from both the classifiers, CNN and DeepMaxout. The performance of the proposed work is validated and compared over the conventional methods in terms of accuracy, precision, and other measures.