Research And Application Of Data Mining Technologies In XML-Based Electronic Patient Record

Hospital Information System (HIS) is gradually put in use in each big hospital;at the same time, the application of Electronic Patient Record (EPR) and the digitization of medical equipment and instrument cause the unceasing inflation of the information in hospital database. However, at present the processing of the majority of hospitals to database only belongs to the low end operations, which are the input, the revision, the inquiry, the deletion and so on. These operations lack the data integration and analysis, and are far from the medical decision-making and the automatic gain from knowledge. By data mining (DM) technology, understanding the correlations and the developing regulations contained in every kind of diseases, summing up effect of every kind of treatment project has great value and prospect to the diagnosis and treatment with diseases and to the research of medical science.This thesis carries on the research based on the massive medical records, which are produced by EPR system. Because of the characteristic of medical record which the liberalization and structuralization unifies, XML serialization and counter-serialization technology based on the two classes about XmlWriter and XmlReader in Microsoft.NET are adopted. This method carries out the storage and the manipulation of patient record. Supported by this approach, the patient records are saved in XML and manipulated in text.Among the methods of data mining, one of the most widespread in application is association rules DM. By the data preprocessing, item dataset that can be mined directly is produced. Through mining association rules with item dataset, many interesting association rules can be found. Based on the above approach, the association analysis with medical data of the diabetes patients in the kidney internal secretion branch is carried on.The main work of this thesis is:(1) Introducing the application about the data mining in the medical domain, analyzing the applied foreground of the data mining on EPR;(2) Introducing the background knowledge of EPR, and the theories of data mining, mainly analyzing the approach to data preprocessing -the key process of data mining;Analyzing the characteristics and process ofmedical data mining, studying the features and preprocessing methods of medical data;(3) Analyzing the technology of EPR system, and researching the methods of data storage and manipulation, which provides the data pool for data mining;Based on the diabetes patient records, extracting the partial information in medical record as the mid-dataset of data mining by preprocessing;(4) Based on the data mining target of EPR system, Posing the association rules analysis as data mining methods;Discussing and analyzing on the two kind of association rule algorithms — Apriori and FP-growth;According to the characteristics of EPR, demonstrating the Apriori algorithm's superiority;(5) Using the Apriori algorithm to mine the association rules between the diabetes patient records, and analyzing the characteristics of the rules, and posing the approach to improve the rules;analyzing the efficiency of Apriori algorithm, and confirming its feasibility.