Enhancing Insulin Dosage Prediction for Diabetes Management Using Multiple-linear Regression with PCA

The optimal dosage of insulin for diabetic patients using MLR and PCA introduces an advanced method for predicting optimal insulin doses for individuals with diabetes by leveraging machine learning techniques. Our approach integrates data preprocessing, feature extraction using Principal Component Analysis (PCA), and the development of multiple linear regression models. We evaluate our models using metrics like Mean Absolute Error (MAE) to demonstrate the limitations of traditional insulin dosage calculation methods. By utilizing historical patient data and real-time monitoring, our approach provides a more personalized and adaptable insulin therapy regimen. Our results indicate that combining PCA with multiple linear regression significantly improves the accuracy of insulin dosage recommendations, offering a promising solution for diabetes management. Additionally, conventional multi-linear regression models often overlook demographic and clinical similarities among patients, relying on generalized data that diminishes predictive accuracy. To address these challenges, By leveraging demographic and clinical similarities, our MLR model predicts optimal insulin dosages for diabetic patients, aiming to maintain effective glucose control and minimize hypoglycemia risk. Moreover, the model forecasts future blood glucose levels, offering proactive alerts to healthcare providers when glucose levels are anticipated to exceed the safe range of 80 to $180 \mathrm{mg} / \mathrm{dL}$. Experimental findings demonstrate that our MLR algorithm significantly supports healthcare professionals in prescribing precise insulin doses, ensuring sustained blood glucose levels within the desired range and thereby improving the overall management of diabetes.