910 likes | 1.91k Views
Chapter 3: SAS Rapid Predictive Modeler. Chapter 3: SAS Rapid Predictive Modeler. Objectives. Present a typical approach to data mining. State the k ey business drivers of SAS Rapid Predictive Modeler.
E N D
Objectives • Present a typical approach to data mining. • State the key business drivers of SAS Rapid Predictive Modeler. • Present an alternative approach to data mining where the business analyst and subject matter expert develops his or her own models. • Describe the key capabilities of SAS Rapid Predictive Modeler.
Churn Case Study Analysis Goal: A telecommunications company wants to decrease the number of churning customers through the development of a churn classification model. Data set: CHURN_RPM Number of rows: 4,708 Number of columns: 15 Contents: account information, call history, equipment and complaint history Targets: TARGET_CHURN (binary)
Churn Case Study: Basics • Throughout this chapter, you work with data in SAS Enterprise Guide and SAS Enterprise Miner to perform fast and accurate modeling with SAS Rapid Predictive Modeler. • Import the CHURN_RPM data. • Build the SAS Rapid Predictive Modeler model in SAS Enterprise Guide. • Score the CHURN_RPM_SCORE data set. • Open the model in SAS Enterprise Miner. • Improve the model.
Data Mining and Predictive AnalyticsConventional Approach Quantitative Modeler / Statistician • Model Development, Deployment, and Management • SAS Enterprise Miner, SAS/STAT • Apply model to specific customer issues (ex. find out customers, which are most likely to churn) • SAS Enterprise Guide, SAS Add-In for Microsoft Office Business Analyst / Subject Matter Expert • Data preparation and data cleansing • SAS Data Integration Studio Database Admin / IT
SAS Rapid Predictive ModelerKey Business Drivers • Need to generate numerous models to support a variety of business problems. • Models need to be developed in a short time-frame using a self-service approach. • Does not have to always rely on a statistician or modeler. • Collaborate to augment, validate, and deploy models.
SAS Rapid Predictive ModelerComplementary Approach • Generate predictive models in a quick, automated fashion • Easy-to-understand reports and charts • Register model in SAS metadata • SAS Enterprise Guide or SAS Add-In for Microsoft Office Business Analyst / Subject Matter Expert Quantitative Modeler / Statistician • Refine model and perform model comparison • Test, validate, and select champion model • Monitor model performance for degradation • SAS Enterprise Miner SAS, Model Manager • Data preparation and data cleansing • SAS Enterprise Data Integration Server Database Admin / IT
SAS Rapid Predictive ModelerPrimary Objectives • Generate predictive models quickly and accurately. • Provide self-sufficiency to business users. • Generate easy-to-understand charts and reports. • Integrate analytics and BI for better decisions.
SAS Rapid Predictive Modeler’s Target Customers • Across all industries • Those dealing with customer-oriented and marketing-analytics-oriented issues • Those who need to generate numerous models to support a variety of business problems: • customer acquisition • up-sell and cross-sell • customer retention • customer churn • Business analysts, subject matter experts, and business professionals with little to no statistical knowledge • Statisticians or data miners who need to develop quick baseline models that address common business issues
What Is SAS Rapid Predictive Modeler? • SAS Rapid Predictive Modeler is a customized task that runs prebuilt SAS Enterprise Miner models. • It is an add-in for SAS Enterprise Guide or SAS Add-In for Microsoft Office. • It requires SAS Enterprise Miner and is included in SAS Enterprise Miner packaging. • It also works with SAS Enterprise Miner for Desktop. • It enables business users, without prior statistical knowledge, to build predictive models quickly and effectively. • Results can be consumed in simple and easy-to-understand charts to make better decisions.
Key Capabilities • You choose from basic, intermediate, or advanced prebuilt methods. • SAS Rapid Predictive Modeler automatically handles outliers, missing values, rare target events, skewed data, variable selection, and model selection. • Analytic results are presented in easy-to-understand business terms: scorecard, lift charts, and listing of key variables in the model. • Analytic experts can further customize and improve models developed in SAS Rapid Predictive Modeler using SAS Enterprise Miner. • Models are registered in SAS metadata to • automate the execution of score code • ease deployment to other systems.
Objectives • Give an overview of the SAS Rapid Predictive Modeler process.
SAS Rapid Predictive Modeler Modeling Process: Overview Open SAS Enterprise Guide or Microsoft Excel. Invoke the SAS Rapid Predictive Modeler task. Select the data to model. Define modeling roles (done automatically if variables are aptly named – for example, target_churn). Run. Review results. (You can save and share them.) (optional) Save task to a SAS Enterprise Miner project. (optional) Register the model in SAS metadata.
Invoke SAS Rapid Predictive Modeler TasksData MiningRapid Predictive Modeler
Invoking and Running the SAS Rapid Predictive Modeler Task Churn Case Study Task: Invoke and execute the Rapid Predictive Modeler task in SAS Enterprise Guide.
Objectives • Give a high-level overview of the SAS Rapid Predictive Modeler model settings.
SAS Rapid Predictive Modeler: Data Panel • Required: • dependent variable (target) • Optional: • Set frequency count. • Set ID. • Exclude input variables. • Edit data and filter. Associate input variables with modeling roles.
SAS Rapid Predictive Modeler: Model Panel • Default: • Basic • Other methods: • Intermediate • Advanced • Other selections: • Decisions and priors Specify the complexity level of the model to build.
SAS Rapid Predictive Modeler: Model Panel Event level Prior probabilities Decision function Decisions and Priors
SAS Rapid Predictive Modeler: Report Panel Select additional features to be included in the model summary report.
Objectives • Review the charts and reports generated as output by SAS Rapid Predictive Modeler.
SAS Rapid Predictive Modeler: Standard Report Output Model Gains Chart
SAS Rapid Predictive Modeler: Standard Report Output ROC Chart
SAS Rapid Predictive Modeler: Standard Report Output Scorecard
SAS Rapid Predictive Modeler: Standard Report Output Project Information
SAS Rapid Predictive Modeler: Optional Report Output Model Summarization
SAS Rapid Predictive Modeler: Optional Report Output Variable Ranking
SAS Rapid Predictive Modeler: Optional Report Output Crosstabulations
SAS Rapid Predictive Modeler: Optional Report Output Classification Matrix
SAS Rapid Predictive Modeler: Optional Report Output Fit Statistics
SAS Rapid Predictive Modeler: Optional Report Output Cumulative Lift Plot
SAS Rapid Predictive Modeler: Optional Report Output * Only available with intermediate or advanced methods Model Comparison*
Objectives • Demonstrate how SAS Enterprise Miner project data from an RPM model can be saved for later inspection and refinement.
SAS Rapid Predictive Modeler: Options Panel Save SAS Enterprise Miner project data from your SAS Rapid Predictive Modeler model.
Objectives • Show how a SAS Rapid Predictive Modeler model can be registered to the SAS Metadata Repository and explain why this might be necessary.
Register the SAS Rapid Predictive Modeler Model • Use Cases: • Import and score using the Model Scoring task in SAS Enterprise Guide. • Import into SAS Enterprise Miner using the Model Import node for integrated model comparison. • Import into SAS Model Manager for champion/challenger model management. • Import into SAS Data Integration Studio to score with mining results transformation. • Publish as a scoring function for Teradata, Netezza, or IBM DB2. Register the model to the SAS Metadata Repository.
Objectives • Demonstrate how a new data set can be scored with SAS Rapid Predictive Modeler. • Discuss the steps of the model scoring task.