T
W
T
F
S
S
M
T
W
T
F
S
1
2
3
4
5
6
7
1
2
3
4
1
2
1
2
3
4
5
6
8
9
10 11 12 13 14
5
6
7
8
9 10 11
3
4
5
6
7
8
9
7
8
9
10 11 12 13
15 16
17 18 19 20 21
12 13 14 15
16 17 18
10 11
12 13 14 15
16
14 15 16
17 18 19 20
22 23
24 25 26 27 28
19 20 21 22
23 24 25
17 18
19 20 21 22
23
21 22 23
24 25 26 27
29 30
31
26 27 28 29
30
24 25
26 27 28 29
30
28 29 30
31
Prediction
Input variables
Target variable
Figure 3.4 Profiling and prediction differ only in the time frames of the input and target variables.
Profiling
Profiling is a familiar approach to many problems. It need not involve any sophisticated data analysis. Surveys, for instance, are one common method of building customer profiles. Surveys reveal what customers and prospects look like, or at least the way survey responders answer questions.
Profiles are often based on demographic variables, such as geographic location, gender, and age. Since advertising is sold according to these same variables, demographic profiles can turn directly into media strategies. Simple profiles are used to set insurance premiums. A 17-year-old male pays more for car insurance than a 60-year-old female. Similarly, the application form for a simple term life insurance policy asks about age, sex, and smoking—and not much more.
Powerful though it is, profiling has serious limitations. One is the inability to distinguish cause and effect. So long as the profiling is based on familiar demographic variables, this is not noticeable. If men buy more beer than women, we do not have to wonder whether beer drinking might be the cause
470643 c03.qxd 3/8/04 11:09 AM Page 54
54
Chapter 3
of maleness. It seems safe to assume that the link is from men to beer and not vice versa.
With behavioral data, the direction of causality is not always so clear. Consider a couple of actual examples from real data mining projects:
■■
People who have purchased certificates of deposit (CDs) have little or no money in their savings accounts.
■■
Customers who use voice mail make a lot of short calls to their own number.
Not keeping money in a savings account is a common behavior of CD holders, just as being male is a common feature of beer drinkers. Beer companies seek out males to market their product, so should banks seek out people with no money in savings in order to sell them certificates of deposit? Probably not! Presumably, the CD holders have no money in their savings accounts because they used that money to buy CDs. A more common reason for not having money in a savings account is not having any money, and people with no money are not likely to purchase certificates of deposit. Similarly, the voice mail users call their own number so much because in this particular system that is one way to check voice mail. The pattern is useless for finding prospective users.
Prediction
Profiling uses data from the past to describe what happened in the past. Prediction goes one step further. Prediction uses data from the past to predict what is likely to happen in the future. This is a more powerful use of data. While the correlation between low savings balances and CD ownership may not be useful in a profile of CD holders, it is likely that having a high savings balance is (in combination with other indicators) a predictor of future CD purchases.
Building a predictive model requires separation in time between the model inputs or predictors and the model output, the thing to be predicted. If this separation is not maintained, the model will not work. This is one example of why it is important to follow a sound data mining methodology.
The Methodology
The data mining methodology has 11 steps.
1. Translate the business problem into a data mining problem.
2. Select appropriate data.
3. Get to know the data.
4. Create a model set.
5. Fix problems with the data.
470643 c03.qxd 3/8/04 11:09 AM Page 55
Data Mining Methodology and Best Practices
55
6. Transform data to bring information to the surface.
7. Build models.
8. Asses models.
9. Deploy models.
10. Assess results.
11. Begin again.
As shown in Figure 3.5, the data mining process is best thought of as a set of nested loops rather than a straight line. The steps do have a natural order, but it is not necessary or even desirable to completely finish with one before moving on to the next. And things learned in later steps will cause earlier ones to be revisited.
1
Translate the
business problem
into a data mining
problem.
2
Select appropriate
data.
3
10
Get to know
Assess results.
the data.
4
9
Create a model set.
Deploy models.
5
8
Fix problems with
Assess models.
the data.
6
7
Transform data.
Build models.
Figure 3.5 Data mining is not a linear process.
470643 c03.qxd 3/8/04 11:09 AM Page 56
56
Chapter 3
Step One: Translate the Business Problem
into a Data Mining Problem
A favorite scene from Alice in Wonderland is the passage where Alice asks the Cheshire cat for directions:
“Would you tell me, please, which way I ought to go from here?”
“That depends a good deal on where you want to get to,” said the Cat.
“I don’t much care where—” said Alice.
“Then it doesn’t matter which way you go,” said the Cat.
“—so long as I get somewhere,” Alice added as an explanation.
“Oh, you’re sure to do that,” said the Cat, “if you only walk long enough.”
The Cheshire cat might have added that without some way of recognizing the destination, you can never tell whether you have walked long enough! The proper destination for a data mining project is the solution of a well-defined business problem. Data mining goals for a particular project should not be stated in broad, general terms, such as:
■■
Gaining insight into customer behavior
■■
Discovering meaningful patterns in data
■■
Learning something interesting
These are all worthy goals, but even when they have been achieved, they are hard to measure. Projects that are hard to measure are hard to put a value on.
Wherever possible, the broad, general goals should be broken down into more specific ones to make it easier to monitor progress in achieving them. Gaining insight into customer behavior might turn into concrete goals:
■■
Identify customers who are unlikely to renew their subscriptions.
■■
Design a calling plan that will reduce churn for home-based business customers.
■■
Rank order all customers based on propensity to ski.
■■
List products whose sales are at risk if we discontinue wine and beer sales.
Not only are these concrete goals easier to monitor, they are easier to translate into data mining problems as well.
What Does a Data Mining Problem Look Like?
To translate a business problem into a data mining problem, it should be reformulated as one of the six data mining tasks introduced in Chapter One:
470643 c03.qxd 3/8/04 11:09 AM Page 57
Data Mining Methodology and Best Practices
57
■■ Classification
■■ Estimation
■■ Prediction
■■ Affinity Grouping
■■ Clustering
■■ Description and Profiling
These are the tasks that can be accomplished with the data mining techniques described in this book (though no single data mining tool or technique is equally applicable to all tasks).
The first three tasks, classification, estimation, and prediction are examples of directed data mining. Affinity grouping and clustering are examples of undirected data mining. Profiling may be either directed or undirected. In directed data mining there is always a target variable—something to be classified, estimated, or predicted. The process of building a classifier starts with a predefined set of classes and examples of records that have already been correctly classified. Similarly, the process of building an estimator starts with historical data where the values of the target variable are already known. The modeling task is to find rules that explain the known values of the target variable.
In undirected data mining, there is no target variable. The data mining task is to find overall patterns that are not tied to any one variable. The most common form of undirected data mining is clustering, which finds groups of similar records without any instructions about which variables should be considered as most important. Undirected data mining is descriptive by nature, so undirected data mining techniques are often used for profiling, but directed techniques such as decision trees are also very useful for building profiles. In the machine learning literature, directed data mining is called supervised learning and undirected data mining is called unsupervised learning.
How Will the Results Be Used?