Data Mining Technology In The Prediction Of Patient Research And Realization

The thesis is achieved mainly during the course of studying the project" The application study and realization of data mining technology on relation rule and artificial neural network on prediction of kinds of patients quantities in JiangSu Tumour Hospital Institute", which is supported by the college level of problem of NanTong Vocational College. Comprehend knowledge and experience of domain experts, select relation rule to analyze the influence factors qualitatively, and then quantitatively predicts the quantities of kinds of patients status by artificial neural network technology, and finally build up a platform of predicting, "Hospital Institute Data Mining (HIDM)" , for the quantities of kinds of patients Status. The main content is as following"It separately presents application in HIDM by relation rule analysis and the technology JiangSu Tumour Hospital Institute artificial neural network. At aspect of study of relation rule analysis, it mines the relation between the influence factors by Apriori algorithm and output after washing and choosing the productive data; at aspect of the technology of artificial neural network, data preparation and neural network's parameter in data mining is discussed mainly. To the existed problem of standard BP arithmetic inefficient and easily getting into local minimum, an improved BP arithmetic is put forward in this paper. And the adjusted frequency and study efficiency of this arithmetic is advanced in stress. Paper has reasoned it not only in theories, but also in experiments using the math logic problem, "XOR" problem. Experiments results show that the improved BP arithmetic has shorted the study time, improved the study efficiency and averted from getting into local minimum in some degree.Then, on the base of above study of theory and methods, this paper researches the achievement and the functions of kinds of modules of platform (HIDM)。Through using the improved BP arithmetic in HIDM, besides the quantities of kinds of patients status are predicted by HIDM, then the result is analyzed. On the whole, the result of testing the software indicates that this platform is capable of achieving the prediction function in data mining.