Leveraging Keras for Early Detection of Diabetes: A Multi-Layer Perceptron Model

Diabetes is a prolonged ailment with serious health hazards, proving itself fatal by taking into its claws around 450 million people globally and causing up to 5 million casualties annually. Type 1 diabetes arises from inadequate insulin production, while early diagnosis often reverses Type 2 diabetes due to insulin resistance resulting from obesity, poor diets, and unhealthy eating habits. The paper aims to design a multilayered perception (MLP) model using Keras for early detection of diabetes by utilizing patient data such as age and blood insulin levels to prompt early diagnosis and treatment. The data base includes parameters like glucose level, blood pressure, skin thickness, insulin levels, BMI, diabetes pedigree function, age, pregnancy period, and diabetes outcome. Handling missing values, standardizing data, and splitting it into training, validation, and testing contribute to data processing. A MLP model is designed with Keras, trained over 200 epochs with the Adam optimizer, and has a binary cross-entropy loss function. We evaluate the model's performance using metrics such as accuracy, confusion matrix, ROC curve, and feature importance. The MLP has approximately 77% accuracy on testing. Key parameters are glucose level, BMI, and age. The confusion matrix and ROC plot exhibit the model's diagnostic ability with balanced true positive and true negative rates. The study proposes an AI model with remarkable accuracy for predicting diabetes onset. It fairly highlights the potential of machine learning to facilitate early diagnosis and smooth diabetes management. The findings lay the groundwork for future research and extensive applications in the healthcare sector.