Diabetes Prediction Using Hybrid 1D CNN and PeeHole Long Short-Term Memory Algorithm

According to the survey of International Diabetes Federation (IDF) the number of persons with diabetes is continuously increasing. There are currently 3820 million diabetics worldwide, and during the next 15 years, this figure might double. Diabetes can create several life-threatening diseases, including heart attack, renal failure, and stroke. To sustain a healthy life, early diabetes identification is crucial. machine learning (ML) is a approach which automatically learns from experience to improve performance and produce more accurate predictions, that can be used to forecast diabetes. Deep learning is a branch of machine learning, and the human brain inspires both its architecture and the way it works.This manuscript presents a hybrid method that combines peephole LSTM and 1D CNN to improve the accuracy of diabetes prediction. Performance is evaluated based on F1 score, accuracy, specificity, sensitivity, and epoch-wise accuracy and loss.The Pima Indian diabetes dataset is utilized for experimental validation. Experimental results demonstrate high prediction accuracy of 96.74% on training and 75.32% on testing for diabetes prediction.