Overview of Data Mining - Java Data Mining: Strategy, Standard, and Practice - page 15

case, making use of this knowledge is key to realizing a return on
investment (ROI) with data mining. Like gold mining, some corpo-
rate data are rich with patterns and insights, others have nothing.
Unlike some gold mining, however, data mining doesn't have “veins”
of knowledge waiting to be pulled out. Rather, the knowledge is dis-
persed in the data, waiting to be discovered by various data mining
techniques.
Since the costs can be high in the exploration and removal of gold from the
hard rock mines, large companies are created in order to raise the money
necessary for the development of the mines, rather than the solitary individual or
small group associated with placer mining.
The machinery brought to bear in data mining, of course, is
computer hardware, software algorithms, and often experienced
data analysts. Traditionally, companies specialized in their ability to
mine certain domains of data, using certain techniques; for exam-
ple, the retail domain for customer segmentation and response
modeling, banking for fraud detection and credit scoring, genomics
for cancer cell similarity analysis, or homeland security for text doc-
ument analysis. Other companies cover a much broader range of
domains and techniques. Still, the image of statisticians or data
mining experts in the back room working creative magic with the
data and producing mind-boggling results is prevalent.
Mining for gold is only worthwhile financially where there is a significant
concentration of it found in ore. The fixed price of gold in 1934 increased
from $20.67 U.S. to $35 U.S. per troy ounce. This price remained fixed until 1968
which discouraged hard rock mining for gold because increased inflation (which
raised the cost of mining production) prevented the mining companies from making
a profit.
The price of gold can be likened to ROI in business. An IDC
report [IDC 2003] shows the median ROI for advanced analytics
projects, such data mining, to be 145 percent. This makes an
investment in such projects a worthwhile venture in general. It is
when competition becomes so fierce, and margins so slim, that not
leveraging data mining becomes a practice dangerous to corporate
survival. Like gold mining, data mining results may miss ROI tar-
gets for numerous reasons: the quality of the raw material (data)
from which knowledge is to be extracted, using the wrong tool(s),
applying the wrong technique to the problem, the skill of the indi-
viduals performing the mining, the inability to use the mining
results effectively in the business process, and so on. This can