A Machine learning approach to reduce the diabetes patient's readmission risk using a novel preprocessing technique

High levels of glucose in the blood are the primary indicator of diabetes. Due to the fact that diabetes can cause several complications, many diabetics are hospitalized multiple times. The quality of treatment that is delivered and the overall cost of medical services are impacted when patients spend unnecessary time in the hospital. The healthcare systems inevitably capture data from clinical patient contacts; a machine learning based approach seems suited for addressing this issue. This paper includes one hundred thousand medical records for seventy thousand diabetic patients from 130 institutions in the United States (United States of America). Using a machine learning technique to process this massive dataset is time-consuming. Here, a unique preprocessing approach is demonstrated that reduces the 55 characteristics of the United States diabetes dataset to 21, with additional machine learning techniques used to predict early readmission. We employed machine learning (ML)-algorithms such as K-Nearest Neighbor (KNN), support vector machines (SVM), decision trees, random forests (RF), logistic regression (LR), gradient boosting (GB), and Gaussian naive bayes (GaussianNB) to evaluate the accuracy of a successful data engineering technique. Gradient Boosting, which beat the other machine learning methods, achieved 77.4 % accuracy in the end.