Predictive Modelling of Diabetes Complications: Insights from Binary Classifier on Chronic Diabetic Mellitus

Chronic diabetes Mellitus (DM) leads to severe consequences such as nephropathy, neuropathy, and retinopathy. Various learning algorithms are employed to forecast and categorize eight issues, encompassing metabolic syndrome and dyslipidemia. Machine learning is revolutionizing various domains by enabling algorithms to extract models from data. The main objective of this proposed study is to utilize binary classifiers to predict and categorize eight problems associated with DM. For this study we have used the dataset obtained from the Jeevandeep Hospital and Research Centre (JHRC) in Baripada, Mayurbhanj, Odisha, India. It includes 1296 records and has nine input variables and eight output categories that indicate different issues. Multiple models were constructed to address problems like Nephropathy, neuropathy, dyslipidemia, diabetic foot, metabolic syndrome, retinopathy, hypertension, and obesity. Different preprocessing operations on missing data, unbalanced data, and feature selection are employed. Binary classifiers were developed for each type of complication, with the number of records varied for each. Subsequently, an evaluation was conducted. The models achieved a maximum of 95.7% accuracy and 97.9% F1 score. The study demonstrates the effectiveness of using selected features to build accurate classifiers for predicting diabetes complications.