Improving Diabetes Classification Accuracy by Incorporating External Factors in a Machine Learning Framework

Diabetes is a chronic condition that significantly increases the risk of developing serious health complications, including heart disease, renal failure, vision impairment, and nerve damage. Early prediction and accurate diagnosis are crucial to preventing these complications. In current clinical practices, a variety of tests are used to collect data for diagnosing diabetes. However, the large volume of available datasets makes it difficult to efficiently extract meaningful insights, often limiting the accuracy of predictions. Machine learning (ML) algorithms have shown promise in analyzing these complex datasets to uncover hidden patterns and predict diabetes-related health risks. This study proposes a novel diabetes prediction model that leverages a Random Forest Classifier to improve classification accuracy. Traditional factors such as age, body mass index (BMI), glucose levels, and insulin resistance are considered, but the study introduces an innovative approach by incorporating external factors (such as lifestyle and socio-economic factors) that could influence diabetes progression. An enforced pipeline model is used in this study, streamlining data preprocessing, feature selection, and algorithm optimization, which results in a more precise prediction framework. This approach brings the prediction accuracy for diabetes classification closer to 77%, a significant improvement over current models.