An Ensemble Approach for the Prediction of Diabetes

Abstract

Diabetes is a very common disease in the world. If diabetes is detected in the early stage, it can be cured easily. Several machine learning techniques are available to predict diabetes in an earlier stage using a data set. This paper presents a review of several machine learning-based methods to predict diabetes. This paper provides the comparative analysis of Naive Bayes, ANN, SVM, KNN, Random Forest, LSTM, CNN, BLSTM, an ensemble of CNN and LSTM, and an ensemble of CNN and BLSTM to predict diabetes. This paper proposed an ensemble approach of CNN and LSTM to predict diabetes and provides an accuracy of 97.14%, precision of 97.30%, recall of 96.30%, F1-score of 96.79%, and AUC value of 0.97. The comparative analysis shows that the performance of the proposed approach is better in comparison to Naive Bayes, ANN, SVM, KNN, Random Forest, LSTM, CNN, BLSTM, ensemble of CNN and LSTM, and ensemble of CNN and BLSTM for the prediction of diabetes.