The Application Of Data Mining In Clinic Laboratory Information System Based On Association Rules And Clustering

Clinic Laboratory Information System (CLIS) is an important module of Hospital Information System. It bases on the database, connects the computer with instrument, with the help of network. It imitates the workflow of clinic laboratory dept. It realizes electronic of laboratory information and automatic of management of laboratory information.The application of CLIS produces large amount of data, which makes it an emergency work to mine the data to get useful knowledge.This paper expatiates the course of data pretreatment, data mining and knowledge discovery, based on two series of data, which are glucose enduring test and blood corpuscle counting. It discusses the application of association rule and clustering analysis in practice.This paper emphasizes on the pretreatment and associate rule mining on the help of diagnosis of diabetes.The data of glucose enduring test comes from basic information of patients, which come from HIS, and glucose enduring test data, which come from LIS, and other information. We use SQLSERVER DTS to induce the data into data ware. We make platform with SQLSERVER ANALYSIS to construct mutidimension data cube.We use cupidity algorithm to stipulate dimension, which is to find useful property from beginning. We use entropy increment technology to find related properties.Associate rule mining is an important application of data mining. We adopt the classic Apriori algorithm, and induce the concept of interest degree to evaluate the interest of the rule.which can be enforced asl-P(B)int erest{A => E) =(l-P(A))x(\-P(AnB))It is more complete and more meaningful. The flow of algorithm is discussed in detail. We will find rules with inhibition like follows:P1(X,Y1)aP2(X,Y2)a...aPi(X,Y!) => have _disease(X,diabetes)After the algorithm completed in Matlab6.5, we get a series of interesting rules, for example:High blood glucose consistence => diabetes High 2hour blood glucose consistence => diabetesThese have the practice meaning of diagnosis of diabetes.In the analysis of blood corpuscle counting, GRAN% and LYM% are thought meaningful:High GRAN% and low LYM% indicates bacteria infection, and low GRAN% and high LYM% indicates virus infection. This paper analyzes blood corpuscle data to prove its clinic meaning, with means based on model.The flow of algorithm is discussed in detail.At last we draw a scatter map with cluster center juk and 3crk ellipse to present the result, which indicate that the data set can be classified into three classes, which present bacteria infections, normal and virus infections.Data mining is an alternate progress. With the advance of user's request, new methods would be induced in. It is believed that the research of this aspect will be deeper and wider.