A predictive risk level classification of diabetic patients using deep learning modified neural network

Abstract

In health care firm, data mining (DM) has an effectual role in predicting the diseases. Today, diabetes is the chief global health issue. Several algorithms are introduced for predicting the diabetes disease and its accuracy estimation. Yet, there is no effectual algorithm for providing the severity of diabetes in respect of ratio which interprets the impact of diabetes on different organs of the human body. To overcome such drawbacks, predictive and risk level classification of diabetes patients using DLMNN and Naïve Bayes (NB) classification methods is system model. This system model system comprises 2 phases namely, phase-1: diabetic disease prediction model, and phase-2: risk analysis. In phase-1, the patient data are taken as of the dataset. Then, from this patient dataset repeated data are removed using HDFS Map Reduce (). Next, as the preprocessing stage, the missing attributes are replaced by averaging the considered data. After that, from the preprocessed data the disease is predicted using DLMNN classification method which results in obtaining the diabetic patient data. Then, the diabetic patient data are sent to phase-2. In phase 2, the missing attributes are replaced using the same average method. Next, the patient data is sorted centered on age utilizing recursive K-means clustering algorithm. Finally, the clustered patient data is classified using the NB classifier algorithm. Experiential results contrasted the system model modified deep learning algorithm with the existing IKMC algorithm in rapports of precision, accuracy, F-measure, and recall. The outcomes confirmed that the system model diabetes prediction and analysis model shows better results on considering the existent methods.