AN EFFICIENT DEEP NEURAL NETWORK APPROACH FOR DIABETES PREDICTION

Abstract: Millions of people all over the world endure from the incessant condition of Diabetes. Early detection and action can lower the likelihood of problems and assist avoid or delay its development. Diabetes has been predicted using machine learning algorithms using a variety of characteristics, including demographics, clinical data, and lifestyle factors. Using a mix of patient data, including age, body mass index and more we present an approach based on deep learning to predict the chance of acquiring diabetes. K Nearest Neighbor(KNN), Logistic Regression(LR), Support Vector Machine(SVM), Decision Tree(DT) and Random Forest(RF), Deep Neural Networks (DNN) are some of the algorithms used. Each algorithm's accuracy is calculated along with the model's accuracy. The approach with a high accuracy level is used as the model to predict diabetes. This strategy may help medical professionals make knowledgeable judgements and give patients personalized care. A number of metrics, such as accuracy and F1 score, are used to assess the effectiveness of the suggested model. Using deep learning concepts by training the properties of a deep neural network(DNN), we suggest a method for diagnosing diabetes. with 98.49% prediction accuracy, and 93% F1 Score. The experimental findings show that when using Deep learning approach, the suggested system offers good outcomes. This strategy may help medical professionals make knowledgeable judgements and give patients personalized care.

Keywords: Diabetes, Supervised Learning, DL, Data mining, KNN, SVM, Light GBM, DT, RF, DNN