Design And Implementation Of Visual Medical Data Mining Platform Based On Weka

There is still a big gap in the field of informatization construction of medical industry between our country and the developed country.The current situation of medical terminology standardization and Application,information isolated island between different HIS systems and a large number of historical data with different structures can not be effectively used are holding developing informatization construction of medical industry back.Along with the development of informationization and medical data accumulation,leadership decision making in hospital is much more difficult.Finding valuable disease diagnosis and prevention rules from massive data is becoming more and more difficult.With the development of data mining and visualization technology,it has been widely used in many fields,but it is still very rare in the field of medical data mining.The work of this paper focuses on how to use data mining and visualization technology in medical data mining after extacting or transfer data into the data warehouse.This paper designs data mining and visulization platform based on weka using J2EE technology,MySQL database,and advanced visualization techniques D3.This platform make full use of mature pentaho BI system and mature data mining tool Weka and other advanced technicals to reduce the essential skills of medical data mining researchers in computer technology and the data mining theory.The innovation of this paper is that we enriched the preprocess filter in order to meet the need of medical datamining and we made valuble attempt in the field of data source or data mining visualization.the most important is that we succeeded in transforming different data structure from database into arff format and realized J48 model visualizaion in the web-end.In addition,we design a integration middleware beetween Weka and Pentaho BI based on WebService.The design of software integration middleware between weka and pentaho BI realized the integration of two heterogeneous systems and this integration is much more deeper than original integration approach.Finally,this paper presents the test results of this platform from aspects of performance,correctness,interaction of visualization.The test results show that algorithms provided by our platform are accurate.Our platform performs slightly worse in efficiency but more interactive in terms of visualization than weka.The efficiency problem are caused by network transmission overhead.Also we present you a real application scenario of exploring the relationship between blood routine examination results and fifteen common hematological diseases using our system and the mining result shows that there is a certain relationship between blood and blood diseases diagnosis.