ABSTRACT

Early diabetes prediction is an important component of preventive healthcare because it allows for targeted medicines and lifestyle changes to reduce the risk of diabetes and its associated health concerns, such as neuropathy, kidney disease, and cardiovascular disease. It is a proactive method to controlling and enhancing public health outcomes associated with diabetes. Large datasets can be analysed using powerful computing techniques such as machine learning and deep learning to predict diabetes risk. These models might include a variety of variables, such as clinical data, lifestyle information, and genetic factors. In the proposed research, hybrid Recursive feature elimination with Fisher score (H-RFFS) method is a parameter collection technique is used to eliminate unwanted data form dataset and processed with prediction method. A deep learning model called Recurrent Neural Diabetics Prediction is utilised to predict early diabetics and their types. The input has been taken as sequential data and time-series prediction tasks. Using Recurrent Neural Diabetics Prediction to forecast early diabetes entails analysing sequential health data to estimate a person's risk of developing diabetes in the future. Early diabetic prediction is performed by using data set Mendeley Data Repository. Dataset that includes fasting blood glucose levels, BMI, HbA1c values, weight,diabetes family history, food, exercise habits, and other health-related information. This model should provide improved performance matric such as Accuracy 90%, Precision 90%, Specificity 98 %and Sensitivity 88%. Keywords: Recurrent Neural Diabetics Prediction, Neural Network, hybrid Recursive feature elimination, Fisher score method, dataset
Publication date: 2026/10/01

https://www.ijbpas.com/pdf/2026/October/MS_IJBPAS_2026_10530.pdf

Download PDF
doi.org/10.31032/IJBPAS/2026/15.10.10530