|
What is Data Mining?
Need to know what is data mining and want
some practical suggestions to reduce information management costs?
What is a data mine?
It's not someplace where Captain Picard sends Data with a pickaxe.
Data mining involves looking for patterns of data and making decisions
based on patterns or dependencies.
Frequently, these decisions are marketing related but may also involve
security, fraud and, depending upon the organization, core business
decisions such as risk severity management.
What
does all this mean?
Suppose that a large hotel chain is analyzing sales data across all
chains and notices that room service charges "Spike" on Monday nights.
That is an interesting fact by itself but further analysis highlights
other patterns e.g.
- The room service charges are for sundry items such as
potato chips, dip and other snacks;
- The quantity ordered seems high for one person;
- Soft drinks or other beverages seem to always be
included;
- There are several room service charges to the same
room on the same night;
- Sometimes the food is charged to a different room,
than where the order was delivered, and to a different guest; and
- Lounge revenues seem to drop on Monday nights.
These patterns may lead hotel management to look at how policies may be
impacting revenue e.g.
- They may determine that several people are gathering
in one room and making room charges;
- They may discover that lounge televisions are tuned
to news channels with no volume;
- They may remember that Monday night is football night;
- They may conclude that having "Monday Night"
football specials in the lounge might be a good way to attract more
business to the higher profit lounge area than room service; and
- They may decide to try a change of policy and monitor
results.
Hotel management business intelligence analysts who know the answer to
what is data mining can exploit data patterns in a variety of ways to
help make marketing decisions.
What is data mining for
credit risk?
A credit card company may analyze transactions for severely delinquent
cardholders and notice patterns of use e.g. in the six months prior to
delinquency, cardholders may typically
- Make charges at liquor outlets;
- Use the card to make food purchases;
- Pay the minimum balance;
- Use one card to pay the minimum balance on a second
card; or
- Use the card to make purchases at establishments that
are not consistent with prior use.
Credit risk management might analyze these patterns and discover that a
very high proportion of delinquent or card holder "write-offs" fall
into a similar pattern prior to becoming a major issue. They may decide
to change risk policy e.g.
- Cardholders who miss one, or two monthly payments
might normally receive a request for prompt payment after the second
month; and
- Risk management may decide that cardholders who start
to fall into the patterns identified above should receive an immediate
telephone call to review credit status; or
- They may decide to reduce cardholder credit limits.
What is data mining for
fraud?
Credit card companies sometimes use data mining services to analyze
patterns of cardholder usage. These patterns are used to establish
fraud tolerance thresholds e.g. a credit card that is always used by
the cardholder to make small retail purchases or restaurant purchases
is suddenly used to make on-line purchases of substantial value. This
may be a trigger for the credit card fraud department to temporarily
suspend the card and contact the cardholder to ensure that the charges
are legitimate.
The discovery of good patterns and potentially bad patterns is part of
business intelligence data mining.
What is data mining for
crime prevention?
The USA Patriot Act, enacted in October 2001 provides the regulatory
guidance for many data mining activities such as anti-money laundering.
Banks are required to examine all customer transactions and report
suspicious behavior. They may accomplish this by using a sophisticated
data mining application such as FORTENT. This risk and compliance
application is used to analyze all customer transactions such a credit
card transactions, direct deposits, wire transfers, and other
monetary transactions. Transactions that fall within certain
pre-defined patterns are flagged and a suspicious activity report is
created.
Summary...
Data mining involves looking for patterns of
data and making decisions based on patterns or dependencies.
Frequently, these decisions are marketing related but may also involve
security, fraud and, depending upon the organization, core business
decisions such as risk severity management.
A successful information management discipline requires business
intelligence analysts who understand what is data mining.
|
