CONCURRENT PREDICTION OF DIABETES AND HYPERTENSION USING DEEP LEARNING

Abstract

Early detection of diabetes and hypertension is essential in preventing severe complications caused by the diseases. This study developed a prediction model using Feed-Forward Deep Neural Network architecture to predict the diseases. A custom dataset generated by combining features from PIMA Indian dataset and PPG-BP dataset is used in training the model. It achieved 93% accuracy in predicting the diseases. Precision and recall scores were also noteworthy, with 95.5% and 94% for concurrent prediction respectively. These results highlight the model’s balanced performance and reliability in real-world healthcare applications. The study addressed limitations in existing single-disease prediction models by focusing on concurrent prediction, which captures the interrelated nature of diabetes and hypertension. Transfer learning played a crucial role in enhancing the model’s performance, taking advantage of pre-training of models to overcome challenges like limited labelled datasets and help in making the concurrent prediction possible by sensitizing the model with features relevant for individual disease. This approach reduced computational overhead and improved generalization, making the model practical for deployment in resource-constrained healthcare settings. Feature selection and engineering, driven by Recursive Feature Elimination (RFE) and domain knowledge, ensured the inclusion of the most relevant attributes, further optimizing the model’s predictive accuracy.