Abstract | This study investigated the role of glycosylated hemoglobin (HbAlc) levels as a predictor of type 2 diabetes mellitus (T2DM). A dataset containing clinical data from 403 patients was investigated using the rough sets approach. The dataset contained a range of physiological features typically associated with diagnosing diabetes, including gylcosylated hemoglobin. It has been reported in many studies that HbAlc has high predictive capacity with respect to T2DM. In order to investigate directly the predictive role of HbAlc, a rough sets approach was used, utilising the classification feature of this machine learning approach. The HbAlc feature was discretised over a range of values (4.0-8.0) and the classification accuracy was established for each level. The first result of this study worth noting is that HbAlc was a significant feature associated with T2DM. Further, the results indicated that HbAlc values in the range of 6.0-7.5 produced the highest classification accuracy. Lastly, we present several rules extracted from this dataset that can provide quantitative rules in human readable form that can be verified clinically. |
---|