• Module 1: Modifying and Combining SAS Data Sets
• Modifying a Data set using the SET statement
  Using Data statement to Stack multiple data sets
  Using Data statement to Merge multiple data set
• Module 2: Manipulating and Transforming Data
  Informants and Formats
  Data Set Option (Drop, Keep, Firstobs, Obs)
  PROC STEP
• Module 3: Introduction to Data Mining Using SAS Miner
  Data Mining Methodologies
  Definition, Description and Business Application
  Best Practice for Data Mining
  Nodes Description and Modelling Steps in SAS Miner
• Module 4: Getting Started with Predictive Modelling
  Opening SAS Enterprise Miner
  Creating a New Project in SAS Enterprise Miner
  The SAS Enterprise Miner Window
  Creating a Data Source for Modelling
  Creating a Process Flow Diagram
  Create Data Source from SAS Tables
  Use the Basic Metadata Advisor
  Set Role and Level meta data for data source variables
  Set the Role of the table (raw, scoring, transactional, etc)
• Module 5: Explore and Assess Data Source for Modelling
  Create and interpret plots, including Histograms, Pie charts, Scatter plot,
  Time series, Box plot
  Identify distributions
  Find outlying observations
  Find number (or %) of missing observations
  Find levels of nominal variables
  Explore associations between variables using plots by highlighting and selecting data
  Compare balanced and actual response rates when oversampling has been performed
  Explore data with the STAT EXPLORER node.
  Explore input variable sample statistics
  Browse data set observations (cases)
• Module 6: Modify Data Source During Model Building
  Replace zero values with missing indicators using the REPLACEMENT node
  Use the TRANFORMATION node to be able to correct problems with input data sources, such as variable distribution or outliers.
  Use the IMPUTE node to impute missing values and create missing value indicators
  Reduce the levels of a categorical variable
  Use the FILTER node to remove cases
• Module 7: Describe Key Predictive Modelling Terms and Concept
  Data partitioning: training, validation, test data sets
  Observations (cases), independent (input) variables, dependent (target) variables
  Measurement scales: Interval, ordinal, nominal (categorical), binary variables
  Prediction types: decisions, rankings, estimates
  Dimensionality, redundancy, irrelevancy
  Decision trees, neural networks, regression models
  Model optimization, overfitting, underfitting, model selection

Go the top

• Module 8: Build Predictive Model using Decision Trees
• Explain how decision trees identify split points
  Build Decision Trees in interactive mode
  Change splitting rules
  Explain how missing values can be handled by decision trees
  Assess probability using a decision tree
  Prune decision trees
  Interpret results of the decision tree node, including: trees, leaf statistics, treemaps, score rankings overlay, fit statistics, output, variable importance, subtree assessment plots
  Explore model output (exported) data sets
• Module 9: Build Predictive Model using Regression
  Explain the relationship between target variable and regression technique.
  Explain linear regression.
  Explain logistic regression (Logit link function, maximum likelihood)
  Explain the impact of missing values on regression models.
  Select inputs for regression models using forward, backward, stepwise selection techniques.
  Adjust thresholds for including variables in a model.
  Interpret a logistic regression model using log odds.
  Interpret the results of a REGRESSION node (Output, Fit Statistics, Score
  Ranking Overlay charts).
  Use fit statistics and iteration plots to select the optimum regression model for different decision types.
• Module 10: Predictive Model Assessment
  Explain reasons for oversampling data.
  Adjust prior probabilities.
  Build a profit/loss matrix.
  Add a profit/loss matrix to a predictive model.
  Determine an appropriate value to use for expected profit/loss for primary outcome (from the data, possibly a mean value).
  Optimize models based on expected profit/loss.
  Compare Models suing Model assessment statistics.
  ROC Chart.
  Score Rankings Chart, including (cumulative) % response chart, (cumulative) Lift chart, gains chart.
  Total expected profit.
  Effect of oversampling.
• Module 11: Model Implementation
  Score data sets within SAS Miner
  Configure a data set to be scored in Enterprise Miner
  Use the SCORE node to score new data
  Save scored data to an external location with the SAVE DATA node

Go to the top