A Case-Based Data Mining Platform - Data Mining: Theory, Methodology, Techniques, and Applications - page 30

the processing flows will be quite similar. Based on these facts, when we deal with a
new problem, we can use a similar case's processing flow as template to solve it.
At this time, it is not ready yet to take a past case's processing flow to reuse,
because the issue about how to get a right case at right time is not concerned. This
issue is a problem of similarity-based retrieval. That is, we compare the similarity
scores of new problem with the past cases, and then we select the most similar case as
the right one to help solve the new problem. For this requirement on similarity-based
retrieval, we need further to define some meaningful and comparable attributes to
calculate similarity scores. Generally, these attributes include industry type, problem
type, business objective, data mining goal, and other, which can determine a data
mining case's processing flow at a general level. For simplifying the description, we
use the term of data mining task to enclose these meaningful and comparable
attributes. Data mining task is attached on the data mining system to retrieve similar
data mining cases. It is also the third extension to generic data mining model.
Now, we can illustrate the data mining model that we have extended. As shown in
Figure 1, the central part of this data mining model is a process builder. It retrieves
similar cases based on data mining task, loads data from the data base, calls operators
from operator base, reuses processing flows to generate model(s) for new data mining
problem, and outputs model(s) to model base.

Task

Data Base

Operator
Base

Processing
Flow

Process
Builder

Model
Base

Fig. 1. Extended Data Mining Model for Knowledge Reuse

This data mining model has used the concept of case-based reasoning (CBR).
Case-based reasoning [1] is a sub-field of Artificial Intelligence (AI). It has been
widely used to solve the problems such as configuration, classification, planning,
prediction, and so on [13]. From the perspective of case-based reasoning, this data
mining model has taken knowledge retrieval and knowledge reuse into consideration,
it has also figured out the content of data mining cases. In the next section, we will
have a close look on data mining case.

3 Data Mining Case

From case-based reasoning perspective, a case is a knowledge container [9]. A case
should be defined and represented at an operable level. In this section, we will
introduce data mining case definition and representation.