Performance Analysis of Conventional Machine Learning Algorithms for Diabetic Sensorimotor Polyneuropathy Severity Classification.

摘要: Background: Diabetic peripheral neuropathy (DSPN), a major form of diabetic neuropathy, is complication that arises in long-term patients. Even though the application machine learning (ML) disease diagnosis very common and well-established field research, its (DSPN) using composite scoring techniques like Michigan Neuropathy Screening Instrumentation (MNSI), limited existing literature. Method: In this study, MNSI data were collected from Epidemiology Diabetes Interventions Complications (EDIC) clinical trials. Two different datasets with variable combinations based on results eXtreme Gradient Boosting feature ranking technique used to analyze performance eight conventional ML algorithms. Results: The random forest (RF) classifier outperformed other models for both datasets. However, all showed almost perfect reliability Kappa statistics high correlation between predicted output actual class EDIC patients when six variables considered as inputs. Conclusions: This study suggests RF algorithm-based can help predict DSPN severity which will enhance medical facilities