This research aims to improve the accuracy of overall disease prediction by analyzing the automatic prediction and recommendation of diabetes disease from the electronic health record diabetes dataset. Diabetes data is acquired from patients and is processed utilizing optimal artificial intelligence techniques during the diabetes data recognition process. This research integrated Machine Learning (ML) based approaches to predict diabetes disease features such as Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), Logistic Regression (LR), K-NN, Naive Bayes (NB), and Gradient Boosting (GB). The GB model is proposed to apply diabetes diagnosis to single-class and multiclass classification problems. In the future, we shall incorporate an auto-feature selection method to design the crossed features and select the features for the prediction and classification model. The diabetes dataset of 1145 Pima Indians: The test uses 330 diabetic and 815 non-diabetic participants. An ensemble of gradient boosting was used in the proposed algorithm to achieve an accuracy of 91.23%. As can be seen, the majority vote-based model employs NB, DT, and SVM classifiers, and its accuracy for the diabetes disease dataset is 73.42%, 80.76%, and 82.51%, respectively. Subsequently, the Inclination helping calculation gives the best exactness to diabetes findings compared to the past calculation.