Databases Reference
In-Depth Information
the predictive information from data sources that is not apparently visible.
The data mining process drills through huge volume of data, to discover the
hidden key facts in order to assist in the decision making process.
2
In
other words, data mining discovers patterns of data that can generate new
knowledge for organizational use. This of course entails huge computations,
and therefore, the process must be automated. For ensuring meaningful
results, the foremost requirement is that the data must have been expressed
in a well-understandable format. The first step in data mining is to describe
the data by summarizing its statistical attributes. The data description
alone does not provide any action plan. A predictive model must be built
based on the patterns determined from known results. Then the model is
tested on results outside the original sample of data.
A complete data mining process is depicted in Fig. 6.2 wherein
historical data is used for training the data mining algorithms which is later
evaluated on a subset of the same data set. Later, this learned model is used
for the purpose of prediction in case of any new data. The learning process
is broadly classified as: (i) supervised, and (ii) unsupervised. In supervised
learning, a tutor must help the system in the construction of the model,
by defining classes and providing positive and negative examples of objects
Fig. 6.2.
Data mining process.
