Getting Started with SAS Enterprise Miner 6.1

SAS Documentation

The correct bibliographic citation for this manual is as follows: SAS Institute Inc. 2009. Getting Started with SAS Enterprise Miner TM 6.1. Cary, NC: SAS Institute Inc. Getting Started with SAS Enterprise MinerTM 6.1 Copyright 2009, SAS Institute Inc., Cary, NC, USA 978-1-59994-321-3 All rights reserved. Produced in the United States of America. For a hard-copy book: No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, or otherwise, without the prior written permission of the publisher, SAS Institute Inc. For a Web download or e-book: Your use of this publication shall be governed by the terms established by the vendor at the time you acquire this publication. U.S. Government Restricted Rights Notice. Use, duplication, or disclosure of this software and related documentation by the U.S. government is subject to the Agreement with SAS Institute and the restrictions set forth in FAR 52.227-19 Commercial Computer Software-Restricted Rights (June 1987). SAS Institute Inc., SAS Campus Drive, Cary, North Carolina 27513. 1st electronic book, December 2009 1st printing, December 2009 SAS Publishing provides a complete selection of books and electronic products to help customers use SAS software to its fullest potential. For more information about our e-books, e-learning products, CDs, and hard-copy books, visit the SAS Publishing Web site at support.sas.com/publishing or call 1-800-727-3228. SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. indicates USA registration. Other brand and product names are registered trademarks or trademarks of their respective companies.

Contents

About This Book. v
Chapter 1 Introduction to SAS Enterprise Miner 6.1. 1
What Is SAS Enterprise Miner?. 1
How Does SAS Enterprise Miner Work?. 2
Benefits of Using SAS Enterprise Miner. 3
Accessibility Features of SAS Enterprise Miner 6.1. 3
Getting to Know the Graphical User Interface. 4
Chapter 2 Learning by Example: Building and Running a Process Flow. 7
About the Scenario in This Book. 7
Prerequisites for This Example. 8
Chapter 3 Set Up the Project. 9
About the Tasks That You Will Perform. 9

Create a New Project. 9

Create a Library. 10

Create a Data Source. 11

Create a Diagram and Add the Input Data Node. 13
Chapter 4 Explore the Data and Replace Input Values. 15
About the Tasks That You Will Perform. 15
Generate Descriptive Statistics. 15
Create Exploratory Plots. 18

Partition the Data. 20

Replace Missing Values. 21
Chapter 5 Build Decision Trees. 23
About the Tasks That You Will Perform. 23
Automatically Train and Prune a Decision Tree. 23
Interactively Train a Decision Tree. 25
Chapter 6 Impute and Transform, Build Neural Networks, and Build a

Regression Model. 29

About the Tasks That You Will Perform. 29
Impute Missing Values. 29

Transform Variables. 31

Analyze with a Logistic Regression Model. 33
Analyze with a User-Specified Neural Network Model. 37
Analyze with an Automatically Selected Neural Network Model. 39
Chapter 7 Compare Models and Score New Data. 43
About the Tasks That You Will Perform. 43

Compare Models. 43

Score New Data. 45
Create a Sorted List of Potential Donors. 46
Appendix 1 SAS Enterprise Miner Node Reference. 49

About Nodes. 49

iv Contents Usage Rules for Nodes. 55
Appendix 2 Sample Data Reference. 57
Sample Data Reference. 57

Glossary. 61

Index. 67

About This Book

Audience
This book is intended primarily for users who are new to SAS Enterprise Miner. The documentation assumes familiarity with graphical user interface (GUI) based software applications and basic, but not advanced, knowledge of data mining and statistical modeling principles. Although this knowledge is assumed, users who do not have this knowledge will still be able to complete the example that is described in this book end-to-end. In addition, SAS code is displayed in some result windows that are produced during the course of the example. However, SAS programming knowledge is not necessary to perform any task outlined in this book.

vi About This Book

Chapter 1
Introduction to SAS Enterprise Miner 6.1
What Is SAS Enterprise Miner?. 1 How Does SAS Enterprise Miner Work?. 2 Benefits of Using SAS Enterprise Miner. 3 Accessibility Features of SAS Enterprise Miner 6.1. 3 Overview of Accessibility Features. 3 Exceptions to Standard Keyboard Controls. 4 Other Exceptions to Accessibility Standards. 4 Getting to Know the Graphical User Interface. 4
What Is SAS Enterprise Miner?
SAS Enterprise Miner streamlines the data mining process to create highly accurate predictive and descriptive models based on analysis of vast amounts of data from across an enterprise. Data mining is applicable in a variety of industries and provides methodologies for such diverse business problems as fraud detection, householding, customer retention and attrition, database marketing, market segmentation, risk analysis, affinity analysis, customer satisfaction, bankruptcy prediction, and portfolio analysis. In SAS Enterprise Miner, the data mining process has the following (SEMMA) steps: Sample the data by creating one or more data sets. The sample should be large enough to contain significant information, yet small enough to process. This step includes the use of data preparation tools for data import, merge, append, and filter, as well as statistical sampling techniques. Explore the data by searching for relationships, trends, and anomalies in order to gain understanding and ideas. This step includes the use of tools for statistical reporting and graphical exploration, variable selection methods, and variable clustering. Modify the data by creating, selecting, and transforming the variables to focus the model selection process. This step includes the use of tools for defining transformations, missing value handling, value recoding, and interactive binning. Model the data by using the analytical tools to train a statistical or machine learning model to reliably predict a desired outcome. This step includes the use of techniques such as linear and logistic regression, decision trees, neural networks, partial least squares, LARS and LASSO, nearest neighbor, and importing models defined by other users or even outside SAS Enterprise Miner.

Chapter 1

Assess the data by evaluating the usefulness and reliability of the findings from the data mining process. This step includes the use of tools for comparing models and computing new fit statistics, cutoff analysis, decision support, report generation, and score code management. You might or might not include all of the SEMMA steps in an analysis, and it might be necessary to repeat one or more of the steps several times before you are satisfied with the results. After you have completed the SEMMA steps, you can apply a scoring formula from one or more champion models to new data that might or might not contain the target variable. Scoring new data that is not available at the time of model training is the goal of most data mining problems. Furthermore, advanced visualization tools enable you to quickly and easily examine large amounts of data in multidimensional histograms and to graphically compare modeling results. Scoring new data that is not available at the time of model training is the goal of most data mining problems. SAS Enterprise Miner includes tools for generating and testing complete score code for the entire process flow diagram as SAS Code, C code, and Java code, as well as tools for interactively scoring new data and examining the results. You can register your model to a SAS Metadata Server to share your results with users of applications such as SAS Enterprise Guide and SAS Data Integration Studio that can integrate the score code into reporting and production processes. SAS Model Manager complements the data mining process by providing a structure for managing projects through development, test, and production environments and is fully integrated with SAS Enterprise Miner.

Accessibility Features of SAS Enterprise Miner 6.1
SAS Enterprise Miner 6.1 includes accessibility and compatibility features that improve the usability of the product for users with disabilities, with exceptions noted below. These features are related to accessibility standards for electronic information technology that were adopted by the U.S. Government under Section 508 of the U.S. Rehabilitation Act of 1973, as amended. SAS Enterprise Miner 6.1 conforms to accessibility standards for the Windows platform. For specific information about Windows accessibility features, refer to your operating system's help. If you have questions or concerns about the accessibility of SAS products, send e-mail to accessibility@sas.com.
Exceptions to Standard Keyboard Controls
SAS Enterprise Miner 6.1 uses the same keyboard shortcuts as other Windows applications, with these exceptions: Instead of using the Windows standard, ALT+Spacebar, the system menu can be accessed by using these shortcuts: primary window: Shift+F10+Spacebar secondary window: Shift+F10+Down There is no keyboard equivalent for accessing the Explore window for a data source via the right-click pop-up menu. However, an alternate control is accessible from the View menu There are no keyboard equivalents for these actions: selecting a SAS Server Directory that is a subdirectory lower in the tree than the default folders in the Create New Project Wizard selecting and editing the value of column attributes in the Data Source Wizard maximizing or minimizing the Result window accessing the Expression Builder in the Transform Variable node
Other Exceptions to Accessibility Standards
Other exceptions to the accessibility standards described in Section 508 of the U.S. Rehabilitation Act of 1973 include the following: On-screen indication of the current focus is not well-defined in some dialog boxes, in some menus, and in tables. High contrast color schemes are not universally inherited. SAS Enterprise Miner 6.1 is not fully accessible to assistive technologies: Many controls are not read by JAWS, and the accessible properties of many controls are not surfaced to the Java Accessibility API. Some content in the Data Source Wizard and Library Wizard is not accessible.
Getting to Know the Graphical User Interface
You use the SAS Enterprise Miner GUI to build a process flow diagram that controls your data mining project.
Display 1.2 The SAS Enterprise Miner GUI
1. Toolbar Shortcut Buttons Use the toolbar shortcut buttons to perform common computer functions and frequently used SAS Enterprise Miner operations. Move the mouse pointer over any shortcut button to see the text name. Click on a shortcut button to use it. 2. Project Panel Use the Project Panel to manage and view data sources, diagrams, results, and project users. 3. Properties Panel Use the Properties Panel to view and edit the settings of data sources, diagrams, nodes, and users. 4. Property Help Panel The Property Help Panel displays a short description of any property that you select in the Properties Panel. Extended help can be found from the Help main menu. 5. Toolbar The Toolbar is a graphic set of node icons that you use to build process flow diagrams in the Diagram Workspace. Drag a node icon into the Diagram Workspace to use it. The icon remains in place in the Toolbar, and the node in the Diagram Workspace is ready to be connected and configured for use in the process flow diagram. 6. Diagram Workspace Use the Diagram Workspace to build, edit, run, and save process flow diagrams. In this workspace, you graphically build, order, sequence, and connect the nodes that you use to mine your data and generate reports. 7. Diagram Navigation Toolbar Use the Diagram Navigation Toolbar to organize and navigate the process flow diagram.

Create a New Project

In SAS Enterprise Miner, you store your work in projects. A project can contain multiple process flow diagrams and information that pertains to them.
For organizational purposes, it is a good idea to create a separate project for each major data mining problem that you want to investigate.
To create the project that you will use in this example, complete the following steps: 1. Open SAS Enterprise Miner. 2. In the Welcome to Enterprise Miner window, click New Project. The Create New Project Wizard opens. 3. Proceed through the steps that are outlined in the wizard. Contact your system administrator if you need to be granted directory access or if you are unsure about the details of your site's configuration. a. Select the logical workspace server to use. Click Next. b. Enter Getting Started Charitable Giving Example as the Project Name. The SAS Server Directory is the directory on the server machine in which SAS data sets and other files that are generated by the project will be stored. It is likely that your site is configured in such a way that the default path is appropriate for this example. Click Next. c. The SAS Folder Location is the directory on the server machine in which the project, itself, will be stored. It is likely that your site is configured in such a way that the default path is appropriate for the example project that you are about to create. Click Next. Note:If you complete this example over multiple sessions, then this is the location to which you should navigate after you select Open Project in the Welcome to Enterprise Miner window. d. Click Finish.

Create a Library

In order to access the sample data sets using SAS Enterprise Miner, you must create a SAS library to indicate to SAS the location in which they are stored. When you create a library, you give SAS a shortcut name and pointer to a storage location in your operating environment where you store SAS files. To create a new SAS library for the sample data, complete the following steps: 1. On the File menu, select New Library. The Library Wizard opens. 2. Proceed through the steps that are outlined in the wizard. Contact your system administrator if you need to be granted directory access or if you are unsure about the details of your site's configuration. a. The Create New Library option button is automatically selected. Click Next. b. Enter Donor as the Name. Then enter the Path to the directory on the server machine that contains the sample data that you downloaded from the Web. For example, if the sample data is located on the desktop of the server machine (denoted by C drive), then you could enter C:\Documents and Settings\<username>\Desktop, where <username> is your user name on the server machine. Click Next. c. Click Finish.

Automatically Train and Prune a Decision Tree
Decision tree models are advantageous because they are conceptually easy to understand, yet they readily accommodate nonlinear associations between input variables and one or more target variables. They also handle missing values without the need for imputation. Therefore, you decide to first model the data using decision trees. You will compare decision tree models to other models later in the example. SAS Enterprise Miner enables you to build a decision tree in two ways: automatically and interactively. You will begin by letting SAS Enterprise Miner automatically train and prune a tree. To use the Decision Tree node to automatically train and prune a decision tree, complete the following steps: 1. Select the Model tab on the Toolbar.
2. Select the Decision Tree node icon. Drag the node into the Diagram Workspace. 3. Connect the Replacement node to the Decision Tree node.
4. Select the Decision Tree node. In the Properties Panel, scroll down to view the Train properties: Click on the value of the Maximum Depth splitting rule property, and enter 10. This specification enables SAS Enterprise Miner to train a tree that includes up to ten generations of the root node. The final tree in this example, however, will have fewer generations due to pruning. Click on the value of the Leaf Size node property, and enter 8. This specification constrains the minimum number of training observations in any leaf to eight. Click on the value of the Number of Surrogate Rules node property, and enter 4. This specification enables SAS Enterprise Miner to use up to four surrogate rules in each non-leaf node if the main splitting rule relies on an input whose value is missing. Note:The Assessment Measure subtree property is automatically set to Decision because you defined a profit matrix in Create a Data Source on page 11. Accordingly, the Decision Tree node will build a tree that maximizes profit in the validation data. 5. In the Diagram Workspace, right-click the Decision Tree node, and select Run from the resulting menu. Click Yes in the confirmation window that opens. 6. In the window that appears when processing completes, click Results. The Results window opens. a. On the View menu, select Model English Rules. The English Rules window opens. b. Expand the English Rules window. This window contains the IF-THEN logic that distributes observations into each leaf node of the decision tree.

models. To overcome this obstacle of missing data, you can impute missing values before you fit the models.
It is a particularly good idea to impute missing values before fitting a model that ignores observations with missing values if you plan to compare those models with a decision tree. Model comparison is most appropriate between models that are fit with the same set of observations.
To use the Impute node to impute missing values, complete the following steps: 1. Select the Modify tab on the Toolbar. 2. Select the Impute node icon. Drag the node into the Diagram Workspace. 3. Connect the Replacement node to the Impute node.
4. Select the Impute node. In the Properties Panel, scroll down to view the Train properties: For class variables, click on the value of Default Input Method and select Tree Surrogate from the drop-down menu that appears. For interval variables, click on the value of Default Input Method and select Median from the drop-down menu that appears. The default input method specifies which is the default statistic to use to impute missing values. In this example, the values of missing interval variables are replaced by the median of the nonmissing values. This statistic is less sensitive to extreme values than the mean or midrange and is therefore useful for imputation of missing values from skewed distributions. The values of missing class variables, in this example, are imputed using predicted values from a decision tree. For each class variable, SAS Enterprise Miner builds a decision tree (in this case, potentially using surrogate splitting rules) with that variable as the target and the other input variables as predictors. 5. In the Diagram Workspace, right-click the Impute node, and select Run from the resulting menu. Click Yes in the confirmation window that opens. 6. In the window that appears when processing completes, click OK. Note: In the data that is exported from the Impute node, a new variable is created for each variable for which missing values are imputed. The original variable is not overwritten. Instead, the new variable has the same name as the original variable but is prefaced with IMP_. The original version of each variable also exists in the exported data and has the role Rejected. In this example, SES and URBANICITY have values replaced and then imputed. Therefore, in addition to the original version, each of these variables has a version in the exported data that is prefaced by IMP_REP_.

Transform Variables

Sometimes, input data is more informative on a scale other than that on which it was originally collected. For example, variable transformations can be used to stabilize variance, remove nonlinearity, improve additivity, and counter non-normality. Therefore, for many models, transformations of the input data (either dependent or independent variables) can lead to a better model fit. These transformations can be functions of either a single variable or of more than one variable. To use the Transform Variables node to make variables better suited for logistic regression models and neural networks, complete the following steps: 1. From the Modify tab on the Toolbar, select the Transform Variables node icon. Drag the node into the Diagram Workspace. 2. Connect the Impute node to the Transform Variables node.

4. In the Diagram Workspace, right-click the Variable Selection node, and select Run from the resulting menu. Click Yes in the confirmation window that opens. 5. In the window that appears when processing completes, click Results. The Results window opens. 6. Expand the Variable Selection window.
Examine the table to see which variables were selected. The role for variables that were not selected has been changed to Rejected. Close the Results window. Note: In this example, for variable selection, a forward stepwise least squares regression method was used that maximizes the model R-square value. For more information about this method, see the SAS Enterprise Miner Help. To use the AutoNeural node to search for and train an optimal neural network configuration, complete the following steps: 1. Select the Model tab on the Toolbar. 2. Select the AutoNeural node icon. Drag the node into the Diagram Workspace. 3. Connect the Variable Selection node to the AutoNeural node.
4. Select the AutoNeural node. In the Properties Panel, scroll down to view the Train properties: Click on the value of the model option Architecture and select Cascade from the drop-down menu that appears. This action causes SAS Enterprise Miner to train only cascade network models. Click on the value of the model option Train Action and select Search. This action causes SAS Enterprise Miner to perform a search to find the best of the candidate network models. 5. In the Diagram Workspace, right-click the AutoNeural node, and select Run from the resulting menu. Click Yes in the confirmation window that opens. 6. In the window that appears when processing completes, click Results. The Results window opens. Maximize the Score Rankings Overlay window. From the drop-down menu, select Cumulative Total Expected Profit.
Again, compare this plot to the plots for the other two models. The shape of the curve is similar for this model to that of the curve for both other models. However, the range of cumulative total expected profit is considerably larger on this plot. For example, if you were to solicit the best 40% of the individuals, the total expected profit from the validation data would be around $6250. Soliciting all of the individuals yields a cumulative total expected profit of about $8900. 7. Close the Results window.

Chapter 7

Compare Models and Score New Data
About the Tasks That You Will Perform. 43 Compare Models. 43 Score New Data. 45 Create a Sorted List of Potential Donors. 46
Now that you have five candidate models to use to predict the best target individuals for your mail solicitation, you can compare them to determine a champion model that you will use to score new data. You will perform the following tasks in order to determine which of the individuals in your organization's database to solicit: 1. You will compare models and select a champion model, which according to an evaluation criterion performs best in the validation data. 2. You will create a new data source for a data set that contains scoring data, which has not been used to build any of the models thus far in the process flow and which does not include values of the target variable. You will score this data using the champion model. 3. You will write SAS code to output, based on the scored data, a list of the top potential donors according to the probability of donating and the profitability matrix that you defined in Create a Data Source on page 11.

Compare Models

To use the Model Comparison node to compare the models that you have built in this example and to select one as the champion model, complete the following steps: 1. Select the Utility tab on the Toolbar. 2. Select the Control Point node icon. Drag the node into the Diagram Workspace. 3. Connect all five model nodes to the Control Point node.
Control Point nodes enable you to better organize your process flow diagram. These nodes do not perform calculations; they simply pass data from preceding nodes to subsequent nodes.
4. Select the Assess tab on the Toolbar. 5. Select the Model Comparison node icon. Drag the node into the Diagram Workspace. 6. Connect the Control Point node to the Model Comparison node.
7. In the Diagram Workspace, right-click the Model Comparison node, and select Run from the resulting menu. Click Yes in the confirmation window that opens. 8. In the window that appears when processing completes, click Results. The Results window opens. 9. In the Fit Statistics window, notice that the logistic regression model was selected as the champion model. The champion model has the value Y in the Selected Model column in the Fit Statistics window. In the model selection node, SAS Enterprise Miner selects the champion model based on the value of a single statistic. You can specify which statistic to use for selection in the node Properties Panel. Because you did not change the value of this property, the default statistic was used, which (because a profit matrix is defined) is average profit in the validation data.

AutoNeural
Use the AutoNeural node as an automated tool to help you find optimal configurations for a neural network model. Use the Decision Tree node to fit decision tree models to the data. The implementation includes features that are found in a variety of popular decision tree algorithms such as CHAID, CART, and C4.5. The node supports both automatic and interactive training. When you run the Decision Tree node in automatic mode, it automatically ranks the input variables, based on the strength of their contribution to the tree. This ranking can be used to select variables for use in subsequent modeling. You can override any automatic step with the option to define a splitting rule and prune explicit tools or subtrees. Interactive training enables you to explore and evaluate a large set of trees as you develop them.

Decision Tree

Dmine Regression
Use the Dmine Regression node to compute a forward stepwise least-squares regression model. In each step, an independent variable is selected that contributes maximally to the model R-square value. Use DMNeural node to fit an additive nonlinear model. The additive nonlinear model uses bucketed principal components as inputs to predict a binary or an interval target variable. The algorithm that is used in DMNeural network training was developed to overcome the problems of the common neural networks that are likely to occur especially when the data set contains highly collinear variables. Use the Ensemble node to create new models by combining the posterior probabilities (for class targets) or the predicted values (for interval targets) from multiple predecessor models. One common ensemble approach is to use multiple modeling methods, such as a neural network and a decision tree, to obtain separate models from the same training data set. The component models from the two complementary modeling methods are integrated by the Ensemble node to form the final model solution Gradient boosting creates a series of simple decision trees that together form a single predictive model. Each tree in the series is fit to the residual of the prediction from the earlier trees in the series. Each time the data is used to grow a tree, the accuracy of the tree is computed. The successive samples are adjusted to accommodate previously computed inaccuracies. Because each successive sample is weighted according to the classification accuracy of previous models, this approach is sometimes called stochastic gradient boosting. Boosting is defined for binary, nominal, and interval targets. The LARs node can perform both variable selection and model-fitting tasks. When used for variable selection, the LARs node selects variables in a continuous fashion, where coefficients for each selected variable grow from zero to the variable's least square estimates. With a small modification, you can use LARs to efficiently produce LASSO solutions. Use the MBR node to identify similar cases and to apply information that is obtained from these cases to a new record. The MBR node uses k-nearest neighbor algorithms to categorize or predict observations. Use the Model Import node to import and assess a model that was not created by one of the SAS Enterprise Miner modeling nodes. You can then use the Model Comparison node to compare the user-defined model with one or more models that you developed with a SAS Enterprise Miner modeling node. This process is called integrated assessment. Use the Neural Network node to construct, train, and validate multilayer, feedforward neural networks. By default, the Neural Network node automatically constructs a network that has one hidden layer consisting of three neurons. In general, each input is fully connected to the first hidden layer, each hidden layer is fully connected to the next hidden layer, and the last hidden layer is fully connected to the output. The Neural Network node supports many variations of this general form. The Partial Least Squares node is a tool for modeling continuous and binary targets. This node extracts factors called components or latent vectors, which can be used to explain response variation or predictor variation in the analyzed data.

DMNeural

Ensemble

Gradient Boosting

LARS (Least Angle Regressions)
MBR (MemoryBased Reasoning)

Model Import

Neural Network

Partial Least Squares

Regression
Use the Regression node to fit both linear and logistic regression models to the data. You can use continuous, ordinal, and binary target variables, and you can use both continuous and discrete input variables. The node supports the stepwise, forward, and backward selection methods. Use the Rule Induction node to improve the classification of rare events. The Rule Induction node creates a Rule Induction model that uses split techniques to remove the largest pure split node from the data. Rule Induction also creates binary models for each level of a target variable and ranks the levels from the most rare event to the most common. After all levels of the target variable are modeled, the score code is combined into a SAS DATA step. Use the TwoStage node to build a sequential or concurrent two-stage model for predicting a class variable and an interval target variable at the same time. The interval target variable is usually a value that is associated with a level of the class target.

Rule Induction

TwoStage
Note: These modeling nodes use a directory table facility, called the Model Manager, in which you can store and access models on demand.
Table A1.5 Assess Nodes Description

Cutoff

The Cutoff node provides tabular and graphical information to assist you in determining an appropriate probability cutoff point for decision making with binary target models. The establishment of a cutoff decision point entails the risk of generating false positives and false negatives, but an appropriate use of the Cutoff node can help minimize those risks. You will typically run the node at least twice. In the first run, you obtain all the plots and tables. In subsequent runs, you can change the node properties until an optimal cutoff value is obtained. Use the Decisions node to define target profiles for a target that produces optimal decisions. The decisions are made using a user-specified decision matrix and output from a subsequent modeling procedure. Use the Model Comparison node to compare models and predictions from any of the modeling nodes. The comparison is based on the expected and actual profits or losses that would result from implementing the model. The node produces the charts that help to describe the usefulness of the model. Use the Score node to manage SAS scoring code that is generated from a trained model or models, to save the SAS scoring code to a location on the client computer, and to run the SAS scoring code. Scoring is the generation of predicted values for a data set that might not contain a target variable. Use the Segment Profile node to examine segmented or clustered data and identify factors that differentiate data segments from the population. The node generates various reports that aid in exploring and comparing the distribution of these factors within the segments and population.

End Groups

Ext Demo

Metadata

Reporter

SAS Code

Start Groups
Here are some general rules that govern the placement of nodes in a process flow diagram: The Input Data Source node cannot be preceded by any other node.
All nodes except the Input Data Source and SAS Code nodes must be preceded by a node that exports a data set. The SAS Code node can be defined in any stage of the process flow diagram. It does not require an input data set that is defined in the Input Data Source node. The Model Comparison node must be preceded by one or more modeling nodes. The Score node must be preceded by a node that produces score code. For example, the modeling nodes produce score code. The Ensemble node must be preceded by a modeling node. The Replacement node must follow a node that exports a data set, such as a Data Source, Sample, or Data Partition node.

Appendix 2

Sample Data Reference
The following table lists the variables that are included in the sample data sets donor_raw_data.sas7bdat and donor_score_data.sas7bdat.
Table A2.1 Variable Variables in the Sample Data Sets Description

CARD_PROM_12

number of card promotions sent to the individual by the charitable organization in the last 12 months one of 54 possible cluster codes, which are unique in terms of socioeconomic status, urbanicity, ethnicity, and other demographic characteristics unique identifier of each individual age as of last year's mail solicitation actual or inferred gender this variable is identical to LIFETIME_AVG_GIFT_AMT lifetime average donation (in $) from the individual in response to all card solicitations from the charitable organization based on the period of recency (determined by RECENCY_STATUS_96NK), which is the last 12 months for all groups except L and E, which are 1324 months ago and 2536 months ago, respectively: 1 if one donation in this period, 2 if two donations in this period, 3 if three donations in this period, 4 if four or more donations in this period H if the individual is a homeowner, U if this information is unknown one of 7 possible income level groups based on a number of demographic characteristics

CLUSTER_CODE

CONTROL_NUMBER DONOR_AGE DONOR_GENDER FILE_AVG_GIFT

FILE_CARD_GIFT

FREQUENCY_STATUS_97NK

HOME_OWNER

INCOME_GROUP

Appendix 2

Variable Description

IN_HOUSE

1 if the individual has ever donated to the charitable organization's In House program, 0 if not amount of the most recent donation from the individual to the charitable organization lifetime average donation (in $) from the individual to the charitable organization total number of card promotions sent to the individual by the charitable organization total lifetime donation amount (in $) from the individual to the charitable organization total number of donations from the individual to the charitable organization maximum donation amount from the individual minus minimum donation amount from the individual maximum donation amount (in $) from the individual to the charitable organization minimum donation amount (in $) from the individual to the charitable organization total number of promotions sent to the individual by the charitable organization median home value (in $100) as determined by other input variables median household income (in $100) as determined by other input variables number of months since the first donation from the individual to the charitable organization number of months since the most recent donation from the individual to the charitable organization number of months since the individual has responded to a promotion by the charitable organization number of months that the individual has been in the charitable organization's database total number of known times the donor has responded to a mailed solicitation from a group other than the charitable organization

LAST_GIFT_AMT

LIFETIME_AVG_GIFT_AMT

LIFETIME_CARD_PROM

LIFETIME_GIFT_AMOUNT

LIFETIME_GIFT_COUNT

LIFETIME_GIFT_RANGE

LIFETIME_MAX_GIFT_AMT

LIFETIME_MIN_GIFT_AMT

LIFETIME_PROM

MEDIAN_HOME_VALUE

MEDIAN_HOUSEHOLD_INCOME

MONTHS_SINCE_FIRST_GIFT

DMNeural node 53

Drop node 51

LARS node 53

library

create new 10

Library Wizard 10

logistic regression 33

Market Basket node 50

MBR node 53

Merge node 50

Metadata node 55

missing values 29

Model Comparison node 43, 54

Model Import node 53

MultiPlot node 18, 51

End Groups node 55

Ensemble node 53

Ext Demo node 55

Neural Network node 37, 53

File Import node 49

Partial Least Squares node 53

Path Analysis node 51

Principal Components node 52

process flow diagram

create new 13

project

SAS Code node 46, 55

Score node 54

Segment Profile node 54

SEMMA steps 1

SOM/Kohonen node 51

Start Groups node 55

StatExplore node 15, 51

Regression node 33, 54

Replacement node 21, 52

Reporter node 55

Rule Induction node 54

Rules Builder node 52

Text Miner node 51

Time Series node 50

Transform Variables node 31, 52

TwoStage node 54

sample data location 8

Sample node 50

Variable Clustering node 51
Variable Selection node 51

Your Turn

We welcome your feedback. If you have comments about this book, please send them to yourturn@sas.com. Include the full title and page numbers (if applicable). If you have comments about the software, please send them to suggest@sas.com.

Paper 114-2009

The Next Generation: SAS Enterprise Miner 6.1 Wayne Thompson and David Duling, SAS Institute, Cary, NC

ABSTRACT

SAS Enterprise Miner 6.1 delivers several new product enhancements to empower both business analysts and seasoned data miners to work more efficiently and produce improved results. A file import node is included for easy access to a broad range of input source types. Extended summary statistics for input variables are also generated to aid the analysts in defining variable roles and identify upfront data errors and trends during the data source definition process. The algorithmic suite has been extended to include powerful LARS and LASSO variable selection. Interactive decision-tree users will be able to model multiple targets for multi-objective segmentation strategy building and predictive modeling. The Reporter node includes new graphics for delivering an analysis-ready report journal to business constituents. Data miners can also share models using the new Model Viewer application. A major emphasis has been placed on extended model deployment capabilities to include market-basket scoring, native scoring in Teradata, and optimized scoring code for delivering faster answers.

INTRODUCTION

SAS Enterprise Miner is the premier data mining solution, delivering an abundance of descriptive and predictive modeling tools along with extensive model deployment alternatives. SAS Enterprise Miner 6.1 released in the first quarter of 2009 continues the SAS dedication to data mining productivity and deployment in which customer feedback helps drive a leading research and development staff. SAS Enterprise Miner 6.1 includes a vast range of new features including integration with the SAS 9.2 System, extended data preparation, enhanced modeling capabilities, improved reporting, and new scoring alternatives. This paper provides an overview of the SAS Enterprise Miner 6.1new features and uses an example of mining data about charitable donations to illustrate some of these features.
OVERVIEW OF SAS ENTEPRISE MINER 6.1

SOFTWARE REQUIREMENTS

SAS Enterprise Miner 6.1 requires SAS 9.2 Platform. The SAS 9.2 system is an improved platform for managing and deploying analytical and business intelligence applications for both single-user applications and multi-user enterprises. SAS Enterprise Miner 6.1 contains changes related to the SAS 9.2 system that improve SAS Enterprise Miner installation, security, and administration.

SOFTWARE MIGRATION

SAS Enterprise Miner 5.3 does not operate with SAS 9.2. If you have existing SAS Enterprise Miner 5.3 project information stored in your SAS Metadata Server, the project information is converted from SAS 9.1.3 format to SAS 9.2 format during the SAS 9.2 / Enterprise Miner 6.1 installation. If you have existing SAS Enterprise Miner 5.3 project data folders that are stored on SAS Workspace Servers, the project data folders do not require conversion for use with SAS 9.2 and SAS Enterprise Miner 6.1. All SAS Enterprise Miner 5.3 project data folders, files, tables, views, and catalogs that are stored on SAS Workspace Servers are compatible for use with SAS 9.2 / Enterprise Miner 6.1. In SAS Enterprise Miner 6.1 you can open existing SAS Enterprise Miner 5.3 projects without any manual conversion process. SAS Enterprise Miner 6.1 projects cannot be converted for use with SAS Enterprise Miner 5.3. If you want to upgrade SAS Enterprise Miner 4.3 project data for use with SAS Enterprise Miner 6.1 you can use the Enterprise Miner project conversion macro. The project conversion macro upgrades SAS Enterprise Miner 4.3 project structures to SAS Enterprise Miner 5.3 project structures. SAS Enterprise Miner 6.1 opens Enterprise Miner 5.3 project structures the user creates by the SAS Enterprise Miner Project conversion macro.

PROJECTS

SAS Enterprise Miner 6.1 project information is now stored and managed in the SAS Metadata Folders. SAS Enterprise Miner 6.1 users create projects in a specific folder location. The default location for new SAS Enterprise Miner 6.1 projects is My Folder. The My Folder location is unique for every user and is a private location. When you create a SAS Enterprise Miner 6.1 project, you can accept the default project location or specify a different folder of your own. For example, either individually or a member of a group, you might store mining projects in a common folder where the projects can be shared. You open projects by using a standard Open Project window that displays the SAS metadata folders tree structure by default (Figure 1).
Figure 1. SAS Enterprise Miner Open Projects Window When the SAS Metadata Server is upgraded from SAS 9.1.3 to SAS 9.2, existing SAS Enterprise Miner 5.3 project information that was stored in SAS Metadata Server is migrated to the shared data folder. SAS administrators can view SAS Enteprise Miner 6.1 project information via the SAS Management Console.

MODELS

SAS Enterprise Miner 6.1 models are stored and managed in the SAS Metadata Folders. You register models to a specific folder location. You can now open or import models by using a standard Open window that displays the SAS metadata folders tree structure by default. When a SAS Metadata Server is upgraded from SAS 9.1.3 to SAS 9.2, existing SAS Enterprise Miner 5.3 models that were stored in SAS Metadata Server are migrated to the shared data folder. SAS Administrators can view SAS Enteprise Miner 6.1 model information via the SAS Management Console.

USABILITY

The user interface for SAS Enterprise Miner 6.1 been updated to include quick search code editors, faster dynamic sample generation for creating interactive plots, a totally new interactive decision tree component , and extended model import capabilities.

CODING IMPROVEMENTS

The SAS code editors and text viewers have been enhanced with a quick text search toolbar that highlights and navigates between selected text search results. This is a great aid when searching for text in SAS code, the SAS log, and SAS output listings. You can launch Quick Text Search from the SAS Enterprise Miner 6.1 main menu, or by using as a keyboard shortcut. The Project Start Code Editor window has been modified to include the SAS log. Convenient access to the SAS log helps when you need to debug or modify Enterprise Miner project start code. The Project End Code Editor window has been eliminated.
INTERACTIVE GRAPHIC SAMPLES
Previous versions of SAS Enterprise Miner provided interactive exploratory graphics based on a sample of values in variable list tables. In Enterprise Miner 6.1, the sample table the software uses to generate interactive graphics has been improved to include only the attribute columns that the user selects, plus any additional Target, ID, Frequency, or Cost variables. This reduces the number of columns required to perform interactive graphic sampling and increases the number of rows of data that are available for graphics. In addition, you can perform variable table list sampling for interactive graphics using a sampling algorithm that is stratified by categorical target variables. This change improves the representation of the sample in the presence of skewed data.

MODEL IMPORT AND EXPORT

In SAS Enterprise Miner 6.1 you can register models directly to the SAS metadata folders tree structure. This provides you with more control over the security, access privileges, and organization of models. You can import a registered model into an existing data mining process flow diagram by using the Model Import node. The score code of the imported model is applied to the data in the process flow diagram, generating new model assessment statistics. The Model Repository window has been removed from SAS Enterprise Miner 6.1. The former flat list of registered models has been replaced by a hierarchical view of models in the SAS metadata folders. The Model Import node provides a list of available models. You can select File --> Open Model from the main menu to open a file utility window to browse the SAS Metadata Folders tree structure and choose a model for inspection. You can also use the Model Import tool to navigate the SAS metadata folders tree structure and choose a model for addition to the process flow diagram.

INTERACTIVE TREE

The Tree Desktop Application has also been replaced by an entirely new interactive decision tree component which requires no separate install or documentation. SAS Enterprise Miner 6.1 enables users to invoke the software using Java Webstart. Examples of the new interactive tree component are provided in the Extended Modeling Features section.
DATA MANAGEMENT AND SUMMARIZATION
Preparing representative input data sources is one of the most important and often tedious tasks of data mining. The process of defining and summarizing data sources for analysis in Enterprise Miner 6.1 has been improved tremendously.

FILE IMPORT NODE

The new File Import node provides a quick and easy-to-use tool for non programmers to import external data files into their SAS Enteprise Miner process flows. The File Import node is located on the Sample tab of the SAS Enterprise Miner tool bar. The node supports importing dBase files, Stata, Microsoft Excel.XLS files, SAS.JMP files, Paradox.DB files, SPSS.SAV files, Lotus.WK1,.WK3, and.WK4 files, in addition to tab-delimited.TXT files, comma-delimited.CSV files, and user-defined delimited.DLM files. The data must be located on the SAS Enterprise Miner client machine or in a network location accessible to the SAS Workspace server (Figure 2). The File Import node provides properties for controlling the maximum size of file to be imported along with other useful options, such as the number of rows to skip, the delimiter that separates columns, and whether or not to use the Data Source Advance Advisor for allocating variable roles. You can also preview before importing them to verify that the data structure is consistent and correct with your expectation.
Figure 2. File Import Node Window. You can redefine variable roles and control what column metadata about the variables is displayed in the variables table display. Instead of displaying enormous tables that have many variable attribute columns, you can configure the variables table by selecting only those attributes that are important to their work. The new descriptive statistics are especially helpful for identifying trends and making decisions about how to treat variables downstream in the analysis. In Figure 3 DONOR_AGE, INCOME_GROUP, and WEALTH_RATING have a high percentage of missing values. Many of these interval variables also have large standard deviation and extreme skewness statistics suggesting that you may want to try transforming the data prior to fitting a regression model.
Figure 3. Variables Table Display provides control over what metadata including new descriptive statistics is displayed in columns. You can use the imported data source with other SAS Enteprise Miner nodes to develop an analysis. For example, you can merge the demographics with the aggregated transactional customer data and partition the resulting table into training, validation, and test data sources.

DEFINE DATA SOURCES FROM THE EXPLORER WINDOW
The SAS Explorer window provides convenient access to allocated SAS tables along with SAS Enterprise project data (Figure 4). You can use the Explorer to quickly locate and view table listings or to develop a plot using one of the many interactive graph components. You can now define input data sources to be used in your analysis simply by dragging and dropping a table onto the diagram workspace The Input Data Source wizard automatically opens, guiding you through the process of defining metadata information about the data source, such as variable roles, measurement levels, and data source type. You can specify whether or not to add the data source as an entry to the Data Source folder of the project. The SAS Explorer has also been enhanced to view and edit (when appropriate) catalog entries of the types SOURCE, LOG, OUTPUT, and XML.
Figure 4. SAS Enterprise Miner 6.1 Explorer. After you have defined the master analytical base table containing primarily aggregated customer donation transactions and other core overlay data, you can easily merge it with the demographic data using the Merge node (Figure 5).
Figure 5. Integrate external data sources using the new File Import node and other SAS Enterprise Miner data preparation nodes. The Append node has also been updated for Enterprise Miner 6.1 to combine training, validation, and test data sets into one training data set to calculate summary statistics on. This feature is very useful when the user wants to compare an imported model developed on a full training table versus an Enterprise Miner model developed on partitioned data.
EXTENDED MODELING FEATURES

LAR AND LASSO

Data miners often have training tables that contain several hundred and even thousands of potential predictors. Feature selection is an important task in data mining both to help ultimately develop parsimonious models that are not overly contaminated with collinear effects and also to generalize reasonably well when the model is applied to new data. The new LARS node of SAS Enterprise Miner 6.1 implements model fitting and variable selection for interval targets by using the SAS/STATGLMSELECT procedure. Methods include not only extensions to standard general linear modeling selection methods (forward, backward, and stepwise) but also the newer least absolute shrinkage and selection operator (LASSO) and least angular regression (LAR) methods of Efron, Hastie, Johnston, and Tibshirani. (2004), respectively. The LAR method starts with no effects in the model and adds effects. At the first step the variable chosen to enter the model is the one most correlated with the target. The absolute value of this coefficient grows until a second effect becomes as correlated with the current residual as the effect in the model. This process continues until all variables enter the model or a selection stopping criterion is met. LASSO deletes parameters based on a version of ordinary least squares where the sum of the absolute regression coefficients is constrained. Additional information about the LAR and LASSO methods is provided in Cohen (2006). The LARS node supports selection criteria, such as, Akaikes information criterion, Sawas Bayesian information criterion, and Mallow C(p) statistics. You can also incorporate validation data or use k-fold cross validation for model evaluation. As do other SAS Enterprise Miner modeling nodes, the LARS node generates score. A variety of diagnostic plots are also provided to evaluate the selection process, as shown in Figure 6. One useful plot is the coefficient paths plot, which displays the change in the coefficients at the different steps as variables enter the model. The vertical line corresponds to the optimal model based on the user-defined model selection criterion, which in this case is the averaged square error for the validation data. The Iteration Plot shows the selected step for the optimal model along with the option for you to choose what selection statistic to display on the vertical axis.

Figure 6. Example LARS Node Results You can use the LARS node as a competitor variable reduction method to the Variable Clustering, Variable Selection, and Decision Tree nodes. Customers also asked for the ability to control how selected variable inputs are combined from two or more predecessor nodes into a successor node. The Merge node of SAS Enterprise Miner 6.1 has been updated to enable you to define rules that specify how to combine input variables from multiple predecessor nodes. The ANY rule sets the input to the rejected role if the variable is rejected in any predecessor node. The MAJORITY rule rejects an input if the input is rejected in the majority of the predecessor nodes. The ALL rule rejects an input variable only if it is rejected in all predecessor nodes. In the example process flow in Figure 7, the union of the input variables from the LARS and the Variable Clustering node are passed to the Regression node. This example passes only the selected inputs from the LARS node to another Regression node for integrated model comparison.
Figure 7. The Metadata node supports defining rules to enable users handle how they want to combine candidate variable inputs.
DEVELOPING SEGMENTATION STRATEGIES INTERACTIVELY BY USING SWITCH TARGETS
A switch targets feature has been added to SAS Enterprise Miner 6.1 so you can select a new dependent variable in a tree leaf and make new splits based on the new target. This is a powerful analytical feature for designing decision trees for segmentation strategies. The donations data contains two target variables: the likelihood that a customer will make a donation (TARGET_B with values of 0 or 1) and the dollar donation amount given that the customer responds (TARGET_D). Rather than develop a more classical two-stage predictive model, you may want to develop rule based segmentation strategies that identify dense pockets of high-dollar donors. Midstream during construction of the tree you can switch targets and spit on the second target as shown in Figure 8.
Figure 8. Splitting on multiple targets Essentially the switch target feature provides a very convenient way to evaluate performance based on alternative targets. The segments can be used to define powerful strategies that are easy to understand by business managers. Score code is created for all targets but model assessment is done only for the primary target. The switch target feature complements well the copy-and-paste descendents feature added to SAS Enterprise Miner 5.3.

REPORTER NODE

The Reporter node generates an analysis-ready report which along with SAS Enterprise Miner model packages provides a concise summary of the analysis for archiving and results sharing. The report is generated in PDF and RTF format and contains all information about the variables, functions, parameters, and the graphs displayed in the Node Results windows. In SAS Enterprise Miner 6.1, the Reporter node provides new SAS ODS (Output Delivery System) functions. The new functions create document graphs, process flow diagrams, and analytical plots that match the graphics that are displayed in the SAS Enterprise Miner user interface. The SAS Enterprise Miner 6.1 Reporter node also provides new Decision Tree results plots for use in PDF and RTF documents (Figure 9). In Reporter node output, the properties list for each node tool indicates the property settings that have been changed from their default values. The Reporter results window now contains a standard external file viewer that you can use to view the document that was produced.
Figure 9. Decision Tree Diagram output from the Reporter node.

MODEL DEPLOYMENT

SCORE NODE
The Score node aggregates score code from the process flow diagram to create a single, deployable score code object. In SAS Enterprise Miner 6.1, the Score node scans and manipulates the SAS score code that the process flow diagram generates to eliminate intermediate code that produces terms that are not deployed in the final model function. The internally manipulated code is called optimized score code. The Score node now creates optimized score code by default. The Score node can also output the non-optimized score code for comparison. For example, the Imputation node can add SAS code that creates many new variables, but a subsequent model selection step may keep only a few of the new terms. The optimized code eliminates unused terms that were created by the Imputation node. The optimized code has a major positive impact on scoring and deployment processes. Fewer variables will need to be saved in the score input data sets in operational systems, which can save enterprises large amounts of resources and labor.
SCORING MODELS IN TERADATA
Many SAS customers store their operational in a Teradata Enterprise Data Warehouse (EDW). SAS Enterprise Miner generates SAS score code which coupled with SAS/Access Interface to Teradata can connect to a Teradata server to extract rows to SAS for scoring. Customers requested the ability to score SAS Enteprise Miner models directly in the database to prevent extracting data and to leverage the shared nothing-architecture of Teradata. In the mid 2000s, SAS Enterprise Miner began supporting the Predictive Modeling Markup Language (PMML) for a core set of data mining functions. SAS Enterprise Miner PMML models can be deployed directly in Teradata using the Teradata Analytic Dataset Generator PMML scoring engine. The SAS Scoring Accelerator for Teradata 1.4 is a new product for publishing SAS Enterprise Miner models into Teradata specific scoring functions for execution directly in Teradata. A primary advantage of scoring models using the SAS Scoring Accelerator for Teradata versus PMML is that a larger class of SAS Enterprise Miner data modification and modeling algorithms are supported. The SAS Enterprise Miner 6.1 Score Code Export node exports score files that are used as input to the publishing macro of the SAS Scoring Accelerator for Teradata (Figure 10).

The Score Code Export node is an extension node delivered with the SAS Scoring Accelerator for Teradata media and can be added to your Enterprise Miner 5.3 or 6.1 installations.
Figure 10. SAS Enteprise Miner Process Flow which includes the Score Code Export node.

CONCLUSION

SAS is honored to present SAS Enterprise Miner 6.1 to its loyal and growing user base who are attending this years SAS Global Forum conference. Many new customer enhancements have been added to this release which is expected to tremendously enhance user productivity and result in better predictive models. SAS is dedicated to delivering the most flexible and extensible data mining system to its user community. The authors look forward to seeing you at SAS Global Forum this year and encourage additional feedback and questions about the product. The future of SAS data mining is very promising. We look forward to working with you on the next chapter of SAS Enterprise Miner.

REFERENCES

Cohen, R.A. 2006. Introducing the GLMSELECT Procedure for Model Selection. Proceedings of the 31 SAS Users Group International Conference. Paper 207-31, Cary, NC: SAS Institute Inc. Efron, B., Hastie, T., Johnstone, I., and Tibshirani, R. (2004), Least Angle Regression (with discussion), Annals of Statistics, 32, 407499.
Mangasarian, O.L. and Musicant, D.R (2000), Lagrangian Support Vector Machines, Technical Report 0006, Data Mining Institute, University of Wisconsin, Madison, Wisconsin. Also in Journal of Machine Learning Research 1, March 2001, 161-177. SAS Institute Inc. SAS Scoring Accelerator 1.4 for Teradata: Users Guide. Cary, NC: SAS Institute Inc. 2008 Whats New in SAS Enterprise Miner 6.1 (2009). See http://support.sas.com/documentation/onlinedoc/miner/index.html

ACKNOWLEDGMENTS

The authors express the upmost appreciation to the entire SAS data mining community for helping develop, test and deliver this new release on the SAS 9.2 platform. We also grateful to our customers, partners, and analysts for the excellent feedback that has helped shape many of these new features.

CONTACT INFORMATION

Your comments and questions are valued and encouraged. Contact the authors at: David Duling Development Director SAS Institute Inc. SAS Campus Dr., S6102 Cary, NC 27513 Work Phone: (919) 531-5267 Email: david.duling@sas.com Wayne Thompson Product Manager SAS Institute Inc. SAS Campus Dr., S6100 Cary. NC 27513 Work Phone: (919) 531-6485 Email: wayne.thompson@sas.com

SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. indicates USA registration. Other brand and product names are trademarks of their respective companies.