A Case-Based Data Mining Platform - Data Mining: Theory, Methodology, Techniques, and Applications - page 37

retrieved, they will be display in similar case management window. The user can
view them one by one and select a proper one as template to solve the new
problem. About the proper case, the general situation is that the most similar case
is the proper one.
5. When a proper similar case is selected as template, it will be opened in case
builder window. For the convenience in further description, the processing flow
displayed in case building window of Figure 5 will be used as example to
describe from now on.
6. In case building window, right-click on Data_1 node to invoke data connector.
With data connector, we connect current problem's input data to this platform. On
data connector, the user can set input data's location, name, type, and other
information.
7. Right-click the followed Oper_1 node to invoke operator adapter. From operator
adaptor, the user can first check this operator's usability by viewing its category,
function, and guidelines. If it is not a proper operator, the user needs to delete it
or insert a new operator ahead of it. If it is a proper operator, the user needs to
connect it by setting its path, name, input ID, parameters, and output ID. At
setting the operator's parameters, we can refer the parameter's guideline to see
how to set its value. After setting all the required values of the operator, the user
can execute it.
8. Do in the same way to check and execute the rest operator nodes. A note is that,
between two successive operators, there is a data node. This data node is the output
of former operator and is also the input of successive operator. The user can view,
or save, or export this intermediate data. At the rear part, some intermediate model
will be generated. The user can view it first and then decide to accept or discard it.
When the final model has been generated, the application scenario of model
building is end.
From this application scenario, we can see that the reusable knowledge, either the
whole processing flow, the operator guideline, or the parameter guideline, are very
helpful to solve a new data mining problem. These knowledge are worked as a
supervisor aside of the user. They can eliminate many perplexities for user, such as,
what steps should be taken, what operators should be used, how the parameters should
be set, and so on.
5 Conclusion
Data mining is a complex and time-consuming process. Data mining practice in
industry heavily depends on data mining professionals to provide solutions. In this
paper, we have proposed a case-based data mining platform, which reuses the
knowledge captured in the past data mining cases to solve new similar problems. This
platform is under developing. The XML-based data mining case representation
language has been defined, and the storage bases, the functional modules and the user
interface have been designed. From its application scenario, we can see that this
platform can eliminate many perplexities, such as, what steps should be taken, what