Machine Learning Models For Prediction of Comorbidities of Diabetes Using Convolutional Neural Network

Diabetes is a serious condition that can lead to the development of other diseases. The basic symptoms of diabetes are increased urination, frequent weight loss, and blurred vision. Comorbidities of diabetes are diseases like heart, kidney, and stroke based on nerve infection, which is also increasing a lot nowadays. Various machine learning models like SVM, regression, decision trees, KNN, and the fusion model are used to predict a disease. The existing model only looked at diabetes in conjunction with heart disease, and its prediction accuracy is low. Based on this problem, the proposed method employs the machine learning (ML) technique to accurately predict diabetes in diseases such as the heart, kidney, and stroke. The proposed model predicts diabetes comorbidities using the Convolutional Neural Network (CNN) algorithm. Initially, data are preprocessed by splitting the data in the ratio of 70:30, and they can be classified as training data and testing data. This training data is applied to the CNN model, which is trained by Tensor Flow, and the model can predict the disease with high accuracy. The proposed model predicts diabetes with comorbidities with high accuracy (about 90
^%
) compared to the existing one.