Research on diabetes Prediction and Intelligent Diagnosis System Based on Machine Learning

Abstract

This paper mainly focuses on the diagnosis and prediction of diabetes, and uses machine learning algorithm to study the prediction model of diabetes and intelligent consultation recommendation system. First, based on the public dataset of diabetes in UCL database, logistic regression model was used to screen variables, and four main characteristics were obtained: Pregnancies, Glucose, BMI and DPF. Secondly, four machine learning algorithms, K-nearest neighbor algorithm (KNN), naive Bayesian algorithm, decision tree and support vector machine, are used to establish the prediction model of diabetes. The model results show that the decision tree algorithm and K-nearest neighbor algorithm have high accuracy in specificity and negative case hit rate; However, the ROC curves of K-nearest neighbor algorithm and support vector machine algorithm are overlapped. Therefore, AUC value is further used to compare the prediction results of the two models. Finally, the prediction model based on support vector machine has certain advantages in prediction accuracy. Finally, taking diabetes patients as the object, on the basis of early prediction, using data mining technology, an intelligent diagnosis system for diabetes based on Java and fuzzy reasoning is constructed. The system functions include system user management, user information management, diabetes diagnosis and advice, system management and other functional modules, of which diabetes diagnosis and advice module is the core module of the system. The intelligent diagnosis system can provide basic medical diagnosis and suggestions, thus improving the diagnosis efficiency and becoming an important auxiliary tool for medical diagnosis.