A novel diabetes prediction model using machine learning and enhanced deep neural network

Abstract

Medical data-base knowledge discovery is a distinctive process. Data Mining techniques aids in exploring useful and interesting facts. Diabetes is considered as a challenge in health care sector. Statistics reveals that 8.3% of world s population are predicted as patients affected by diabetes. In U.S, 34 million people are classified as diabetic patients in 2018. It is announced that sixty six million Indians are affected by diabetes. Predicting the diseases in health care industry is effectively done by data mining techniques. Many existing algorithms are used for diabetes disease prediction and estimation of its accuracy. The death rate rises each year and large number of population can be saved from death if their health condition and risk levels related to the diabetes disease is predicted earlier. However, there is a lack of an efficient algorithms for predicting risk levels associated with diabetes disease that deals with diabetes impact on various human organs. This thesis provides three significant contributions to overcome the drawbacks discussed and also to save lives by early prediction of risk levels which helps the physicians to provide the right treatment earlier before it reaches its severity. The solution for the above mentioned issues comprises of three different phases. First phase as diabetes Prediction model, Second phase as Risk analysis and Third phase as early prediction model. In First phase, the redundant data are eliminated using Hadoop distributed file system. The missing attributes are replaced by averaging method as a pre-processing step. Then the disease prediction is done using Deep Learning Modified Neural Network (DLMNN) classification that helps in obtaining input data which is affected by diabetes disease. Optimized weights are obtained using Cuckoo Search Optimization Algorithm (CSOA). The dataset size been shrinked and it results in minimized computation time. In second phase, averaging method is utilized in order to replace missing values from the collected data.