Artificial Intelligence-Based Healthcare Data Analysis Using Multi-perceptron Neural Network (MPNN) Based On Optimal Feature Selection

In the healthcare industry, early identification of diabetes is a critical task to safeguard human life. Many individuals suffer the consequences of undetected diabetes due to the lack of consistent early identification of the disease's nature. The integration of Machine Learning (ML) and artificial intelligence models has emerged as a remarkable advancement in identifying diseases for prompt treatment. However, existing techniques have struggled to prioritize the importance of diabetic features, resulting in lower precision rates and poor prediction accuracy. This research proposes an improved diabetic disease prediction system based on Multi Perceptron Neural Network (MPNN). Initially, preprocessing is conducted to eliminate noise from the diabetic dataset. Subsequently, diabetic margin features are scrutinized to identify crucial features using a Social Spider Optimized Support Scalar Vector (SSOSSV). The selected features are then trained using the MPNN neural network to predict diabetic features. The proposed system efficiently identifies diabetic patients by analyzing feature sets with marginalized medical margins. By the identification, the data is securely processed with blockchain security by utilizing the AES algorithm, increasing round key increments can enhance system security and performance. The results demonstrate the effectiveness of the proposed diabetic prediction model, showcasing higher prediction rates, improved precision, recall rates, F1 measures, and reduced false rates, Higher security and authentication and verification which is all achieved with manageable time complexity. The proposed MPNN method has achieved significant improvement, increasing accuracy to 94.46%, and blockchain technology improves security for diabetic patient records.