Implementing AI Algorithms for Predicting Diabetes Risk in Patients Using Health Informatics Data

Abstract

Diabetes mellitus is a widespread health problem that puts people's health at great risk and puts a lot of stress on healthcare systems around the world. Early detection and forecast of diabetes are very important for managing and preventing the disease. One potential way to improve diabetes risk forecast is to use the power of artificial intelligence (AI) in health information data. The study's goal is to test and apply AI systems that can figure out a person's chance of getting diabetes by looking at a lot of health information, such as demographic, clinical, and lifestyle data. Several AI methods are used in our method, such as logistic regression, decision trees, support vector machines (SVM), and deep learning models like artificial neural networks (ANN). These methods were chosen because they have been shown to work well with complicated, high-dimensional health data. A lot of information from electronic health records (EHRs) is used in the study. This information includes patient details, medical history, lab test results, and lifestyle factors. To make sure the data was correct and useful, it was put through steps like normalization, handling missing numbers, and feature selection. We used a method called stratified k-fold cross-validation to train and test the models. This made sure that the evaluations were accurate and reduced the chance of overfitting. Key performance indicators like F1 score, accuracy, precision, recall, and the area under the receiver operating characteristic curve (AUC-ROC) were used to judge the model's performance. When the algorithms were compared, it was found that deep learning models, especially the ANN, were better at predicting diabetes risk (with an AUC-ROC of 0.92). The feature importance analysis showed that body mass index (BMI), fasting blood glucose levels, age, and a history of diabetes in the family are all strong factors of diabetes risk. We also looked into how AI models can be understood by using methods like SHapley Additive Explanations (SHAP) to show how different traits affect the results of predictions. This ability to be interpreted is very important for getting clinical insights and building trust among healthcare professionals. Our results show that AI systems might be able to correctly guess who will get diabetes, which would allow for early prevention and more personalized treatment plans.