Big Data Predictive Analysis for Type-2 Diabetes Based Heart Disease Using Feature Extraction and Classification by Machine Learning Architectures

Abstract

Machine learning (ML), a branch of AI, enables computers to learn without being explicitly programmed. ML is widely applied in the healthcare industry to forecast a variety of chronic conditions. For improved clinical paths to prevent complications and postpone the onset of diabetes, earlier diabetes prediction is essential. This research propose novel technique in type 2 diabetes based heart disease detection in big data predictive analysis using machine learning method. Input data has been collected as type 2 diabetes and processed for noise removal and dimensionality reduction. Then the processed data features has been extracted for detecting the abnormality of type 2 diabetes using regression model based linear discriminant analysis. The extracted features shows the abnormal type 2 diabetes and for predicting heart disease by classifying the extracted data using VGG-16 Net_gradient NN. Experimental analysis has been carried out in terms of accuracy, precision, recall, F-1 score, RMSE and MAP for various diabetes dataset. Proposed technique attained accuracy of 96%, precision of 67%, recall of 79%, F-1 score of 63%, RMSE of 66% and MAP of 68%.