Diabetes Clinical Data Analysis Based On Data Mining Technology

Data mining, emerging as a interdiscipline, involves the relevant theoretical knowledge of mathematical statistics, machine learning, computational intelligence, database and information retrieval etc.. Data mining applications in many industries for its successful development has provided great impetus. Diabetes since first recorded in 1500 BC, has been troubled by an important human health problem. As information technology in the medical industry more widely, resulting in a large amount of medical information data, and data mining provides a way from which to explore value. It has a significance for medical resource planning, clinical treatment and treatment guidance.In this paper, I analyzed diabetes clinical data using data mining techniques. It is desirable from which to explore the clinical diabetes drug laws, readmission factor, diabetic retinopathy and other knowledge, in order to establish the appropriate classification and decision-making model. In this paper, I complete the following three areas:1. Analyzed the situation of national diabetes present by cluster analysis on the global 219 countries and regions in the country to get the law of the difference between different types of diabetes, the patient’s age, sex and urban and rural distribution. Overall, the rate of patients, witch with low and middle income countries in the world, focus on the 80%. But in this part, the percent of the national health care costs for diabetes only to 20% of the global total treatment costs. Obviously it is an uneven distribution of urban and rural areas. Especially in low-income countries, the prevalence of rural population was significantly higher than in urban sick population.2. By using fuzzy multilayer perceptron as a support vector machine kernel function, and asymmetric triangular membership function, right initialize multilayer perceptron weight and threshold, the formation of fuzzy support vector machine, and as a basis function, classifier using Adaboost algorithm as an integrated learning algorithm to build Adaboost-FSVM model. And applied to data classification Pima Indian diabetic and diabetic retinopathy recognition. In this paper I have compared the SVM, DT, Adaboost-SVM algorithms and found that in two types of diabetes Adaboost-FSVM, witch relevant classification have better effect.3. Analysis the diabetes clinical hospital data, and in data cleansing phase contrast method uses four questions deal with missing data. Then again hospitalized for diabetes patients were analyzed from both sources and overall, followed by digging out the effects of diabetes drugs commonly prescribed agents for change and rehospitalization. Finally, using RBF, SVM, RVM and DT established diabetes readmissions predictive models.