process. Such tools can produce good results, but the expert data
miner may be able to produce superior results through custom
analysis and crafty techniques. In addition, including data mining
results in applications or operational systems has also become sim-
pler as a result of standard interfaces and model representations.
The gold milling process may be broken down into three basic procedures:
Sorting the ore by size
Crushing the rock
Extracting the gold
A simplified data mining process can be broken down into four
Acquiring and preparing the data
Building the model
Assessing model quality and reviewing the model details
Applying the model to new data for predictions or assign-
First, miners raise the ore out of the mine in wheeled carts pushed on rails and
take it down to the mill. The rock fragments are sorted according to size in a
grizzly —a device consisting of a series of spaced bars, rails, or pipes—above a for-
ward moving conveyer belt to a crusher machine.
In large companies, and even some smaller ones, the IT
department's database administrators (DBAs) help to identify
available data from operational systems and data warehouses.
Data is “sorted” according to quality, completeness, and applicabil-
ity to the problem to be solved. Once data is identified, it needs to
be unified by joining different data tables, often into a single table,
or perhaps a set of tables related by a single case identifier or as
part of a star schema.
After secondary washing, a shaker screen filters out fragments of less than
1/2 inch diameter into a fine ore bin, or box. Larger ore fragments are pul-
verized or crushed in the crusher. The fine ore is fed by conveyer belt to a ball mill,
a rotating steel cylinder filled with tumbling steel balls which further crushes the
fragments to a consistency of fine sand or talcum powder. This powder is fed into a
thickener with a cyanide and water solution to create a sludge (a sticky, mudlike