■■ “Account number is NULL” may be synonymous with failure to
respond to a marketing campaign. Only responders opened accounts and were assigned account numbers.
■■ “Date of churn is not NULL” is synonymous with having churned.
Another danger is that the column reflects previous business practices. For instance, the data may show that all customers with call forwarding also have call waiting. This is a result of product bundling; call forwarding is sold in a product bundle that always includes call waiting. Or the data may show that almost all customers reside in the wealthiest areas, because this where customer acquisition campaigns in the past were targeted. This illustrates that data miners need to know historical business practices. Columns synonymous with the targets should be ignored.
T I P An easy way to find columns synonymous with the target is to build decision trees. The decision tree will choose one synonymous variable, which can then be ignored. If the decision tree tool lets you see alternative splits, then all such variables can be found at once.
Model Roles in Modeling
Columns contain data with data types. In addition, columns have roles with respect to the data mining algorithms. Three important roles are: Input columns. These are columns that are used as input into the model.
Target column(s). This column or set of columns is only used when building predictive models. These are what is interesting, such as propensity to buy a particular product, likelihood to respond to an offer, or probability of remaining a customer. When building undirected models, there does not need to be a target.
Ignored columns. These are columns that are not used.
Different tools have different names for these roles. Figure 17.4 shows how a column is removed from consideration in Angoss Knowledge Studio.
470643 c17.qxd 3/8/04 11:29 AM Page 548
548 Chapter 17
Figure 17.4 Angoss Knowledge Studio supports several model roles, such as ignoring a column when building a model.
T I P Ignored columns play a very important role in clustering. Since ignored columns are not used to build the clusters, their distribution in the clusters can be very informative. By ignoring columns such as customer profitability or response flags, we can see how these “ignored” columns are distributed in the clusters. And we might just discover something very interesting about customer profit or responders.
There are some more advanced roles as well, which are used under specific circumstances. Figure 17.5 shows the many model roles available in SAS
Enterprise Miner. These model roles include:
Identification column. These are columns that uniquely identify each row.
In general, these columns are ignored for data mining purposes, but are important for scoring.
Weight column. This is a column that specifies a “weight” to be applied to each row. This is a way of creating a weighted sample by including the weight in the data.
Cost column. The cost column specifies a cost associated with a row. For instance, if we are building a customer retention model, then the “cost”
might include an estimate of each customer’s value. Some tools can use this information to optimize the models that they are building.
The additional model roles available in the tool are specific to SAS Enterprise Miners.
470643 c17.qxd 3/8/04 11:29 AM Page 549
Preparing Data for Mining 549
Figure 17.5 SAS Enterprise Miner has a wide range of available model roles.
Variable Measures
Variables appear in data and have some important properties. Although databases are concerned with the type of variables (and we’ll return to this topic in a moment), data mining is concerned with the measure of variables. It is the measure that determines how the algorithms treat the values. The following measures are important for data mining:
■■ Categorical variables can be compared for equality but there is no meaningful ordering. For example, state abbreviations are categorical. The fact that Alabama is next to Alaska alphabetically does not mean that they are closer to each other than Alabama and Tennessee, which share a geographic border but appear much further apart alphabetically.
■■ Ordered variables can be compared with equality and with greater than and less than. Classroom grades, which range from A to F, are an example of ordered values.
■■ Interval variables are ordered and support the operation of subtraction (although not necessarily any other mathematical operation such as addition and multiplication). Dates and temperatures are examples of intervals.
470643 c17.qxd 3/8/04 11:29 AM Page 550
550 Chapter 17
■■ True numeric variables are interval variables that support addition and other mathematical operations. Monetary amounts and customer
tenure (measured in days) are examples of numeric variables.
The difference between true numerics and intervals is subtle. However, data mining algorithms treat both of these the same way. Also, note that these measures form a hierarchy. Any ordered variable is also categorical, any interval is also categorical, and any numeric is also interval.
There is a difference between measure and data type. A numeric variable, for instance, might represent a coding scheme—say for account status or even for state abbreviations. Although the values look like numbers, they are really categorical. Zip codes are a common example of this phenomenon.
Some algorithms expect variables to be of a certain measure. Statistical regression and neural networks, for instance, expect their inputs to be numeric.
So, if a zip code field is included and stored as a number, then the algorithms treat its values as numeric, generally not a good approach. Decision trees, on the other hand, treat all their inputs as categorical or ordered, even when they are numbers.
Measure is one important property. In practice, variables have associated types in databases and file layouts. The following sections talk about data types and measures in more detail.
Numbers
Numbers usually represent quantities and are good variables for modeling purposes. Numeric quantities have both an ordering (which is used by decision trees) and an ability to perform arithmetic (used by other algorithms such as clustering and neural networks). Sometimes, what looks like a number really represents a code or an ID. In such cases, it is better to treat the number as a categorical value (discussed in the next two sections), since the ordering and arithmetic properties of the numbers may mislead data mining algorithms attempting to find patterns.
There are many different ways to transform numeric quantities. Figure 17.6
illustrates several common methods:
Normalization. The resulting values are made to fall within a certain range, for example, by subtracting the minimum value and dividing by the range. This process does not change the form of the distribution of the values. Normalization can be useful when using techniques that perform mathematical operations such as multiplication directly on the values, such as neural networks and K-means clustering. Decision trees are unaffected by normalization, since the normalization does not change the order of the values.
470643 c17.qxd 3/8/04 11:29 AM Page 551
Preparing Data for Mining 551
Original Data
Normalized to [0, 1]
7,000
1.0
6,000
0.8
5,000
4,000
0.6
3,000
0.4
2,000
0.2
1,000
0
0.0
Time
Time
Standardized
Binned as Deciles
4
10
9
3
8
7
2
6
5
1
Decile 4
3
0
2
1
-1
0
Time
Time
Figure 17.6 Normalization, standardization, and binning are typical ways to transform a numeric variable.
Standardization. This transforms the values into the number of standard deviations from the mean, which gives a good sense of how unexpected the value is. The arithmetic is easy—subtract the average value and divide by the standard deviation. These standardized values are also called z-scores. As with normalization, standardization does not affect the ordering, so it has no effect on decision trees.
Equal-width binning. This transforms the variables into ranges that are fixed in width. The resulting variable has roughly the same distribution as the original variable. However, binning values affects all data mining algorithms.
Equal-height binning. This transforms the variables into n-tiles (such as quintiles or deciles) so that the same number of records falls into each bin. The resulting variable has a uniform distribution.
Perhaps unexpectedly, binning values can improve the performance of data mining algorithms. In the case of neural networks, binning is one of several ways of reducing the influence of outliers, because all outliers are grouped together into the same bin. In the case of decision trees, binned variables may result in child nodes having more equal sizes at high levels of the tree (that is, instead of one child getting 5 percent of the records and the other 95 percent, with the corresponding binned variable one might get 20 percent and the other 80 percent). Although the split on the binned variables is not optimal, subsequent splits may produce better trees.
470643 c17.qxd 3/8/04 11:29 AM Page 552
552 Chapter 17
Dates and Times
Dates and times are the most common examples of interval variables.. These variables are very important, because they introduce the time element into data analysis. Often, the importance of date and time variables is that they provide sequence and timestamp information for other variables, such as cause and resolution of the last complaint call.