Diabetes Complications Prediction Using Different Multi-label Classification Algorithms-MEKA

Abstract

Now-a-days, Diabetes is the most prominent chronic or non-communicable disease (NCD) in India. More Indian families are influenced by this disease due to its increasing occurrence. The Electronic Health Records (EHRs) contain all the clinical related data of patients. The clinical data helps in identifying the patient hidden knowledge like disease identification based on the features and in identifying the correlation between the clinical parameters and the complications that occur out of disease. Data mining algorithms does these tasks. In our work, prediction model is build using different multi-label classification algorithms like Binary Relevance, Label Combination, Pruned Set, RAkEL, and Chained Classifiers have been considered. We can see that the performance of model built using RAkEL and Chained Classifiers are relatively high when compared to Binary Relevance, Least Combination, and Pruned Sets. Model is used to predict diabetes complications from patient records.