Develop A Hybrid Improved Smooth Support Vector with A Modified Pigeon Search Optimization to Detect the Diabetes Mellitus at Early Stage

Diabetes Mellitus (DM) is a typical metabolic disease in which individuals struggle with high blood sugar (i.e., chronic hyperglycemia). It affects most of the body parts, such as the heart, kidneys, eyes, feet, skin, etc. Several competent Diabetes Diagnosis Systems (DDS) exploit different Machine Learning (ML) algorithms to gain valuable insights from the clinical datasets for DDS and disease management. However, trapping into local optimum solution, lack of privacy, missing values in the input dataset, and deficiency of incremental classification are major issues related to conventional ML-based diabetes classification algorithms. The primary objective of this study is to create an effective DM classification model that can reliably identify patient data as normal or diabetic. A hybrid Improved Smooth Support Vector Machine (ISSVM) with a Modified Pigeon Search Optimization (MPSO) algorithm was developed to detect the DM at an earlier stage. The empirical results reveal that the proposed classifiers outperform other related classifiers of existing systems regarding designated performance measures. These proposed algorithms can support clinicians in enabling the secured and accurate classification of DM with better accuracy and other performance measures.