Research And Application On Diabetes Complication With Data Mining Based On OpenEHR

In recent years, many countries have promoted electronic health record (EHR) to solve the problems of medical information sharing and exchange, software engineering development in traditional hospital information system (HIS). The research of chronic diseases based on EHR can not only assist the prevention of chronic diseases in management, but also provide researchers with data analysis. Diabetes have became a serious chronic disease that affect physical health. Many studies of diabetes data mining will cause privacy leakage without data anonymization in the processing of minging and release data. On the other hand, the previous studies only considered the association between diabetes and complications, but ignored the factors associated with diabetes.In this paper, we designed a data mining system for diabetes complications based on OpenEHR platform. To slove the problem of privacy leakage, this paper applied L-diversity method to anonymize dataset in the processing of diabetes data mining. On the basis of anonymization, we applied the entropy minimization method to discrete continuous attributes. In this paper, we also proposed an improved FP-Tree data structure based on frequent priority table. We founded that the FP-Growth algorithm based on improved FP-Tree data structure had improved computational efficiency by comparing with the naive FP-Growth algorithm. Then we proposed a diabetic complication data mining algorithm based on interestingness measure and FP-Growth algorithm by fully considering the influence of relevant factors and analyzing the limitations of support-confidence frame. On the basis of the implementation of above algorithms and modules, we tested all the functions and performances of our system. The results showed the improved FP-Growth algorithm can filtering the negative and weak association rules.