| Summary: | 碩士 === 國立陽明大學 === 公共衛生研究所 === 87 === The technologies on Internet have inspired the informative procedures, and the development of the Database Management Systems has rapidly accumulated a huge amount of the data. As the time goes by, the data quantities may be expanded inappropriately, but the information inside becomes relatively insufficient.
Also in the medical field, because of the development of the Medical Informatics and the digitalization of the NII, there might be also lots of information which is about the patients’ and their medical conditions in medical database. There may exist some useful knowledge behind the data, such that a new point of view, which is applying the data mining technologies into the medical field, has been arisen. It is obvious that there must be able to benefit both mutually, and increase the value of the Medical Informatics.
In this research, we design and implement a Java-Based Data Mining Tool (JBDM tool) utilizing the environment of the Internet. Finally, we apply it on the Medical data. The data mining method we used is adapted from Artificial Intelligence research filed. We use Inductive Learning and choose the C4.5 decision tree algorithm to find the patterns and trend in data, and use it to proceed the prediction for the new cases.
The user can use Java-enabled browser to execute the entire data mining processes, and doesn’t have to worry about what platform he is using. It shows a great mobility. We also express the decision trees structure by using diagram, and make it visualized. We can also simulate the expert’s opinions to change the order of the nodes. About the connection to the database server on Internet, not only make the data resources more extensively, but also can protect the original data in private and safe way via the settings of the authority. This asset is particularly important to the clinical data.
We also try to mine a set of medical data with our system. We use the training data sets to construct the decision tree, then test the external validity with a set of testing data, which outcome has already been known. In our preliminary evaluation, the JBDM tool shows a positive performance.
|
|---|
