Type 2 diabetes data processing with EM and C4.5 algorithm

This research combines EM (expectation maximization) algorithm and C4.5 algorithm together to build a type 2 diabetes data processing system. Taking the advantage of the nearly 14000 items of multi-source, multi-dimension practical dataset, and a series of data mining experiments are designed. With the large quantity of experiments and results analysis, some valuable pathological knowledge of type 2 diabetes was discovered, which includes, the decision tree is almost identical with the list of clinical diabetic risk factors, and the rate of correct recognition for healthy people was 80.9% while for diabetic patients was 92.05%; three new blood glucose threshold 5.85 mmol/l, 5.26 mmol/l and 4.28 mmol/l, The valuable results are good to the cure and macro-control type 2 diabetes.