PR4_2010 - screen shot of your final Diagram as shown on...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
DSCI 4520/5240 – Data-Based Decision Support Systems PR4 Data: DONOR_RAW Analysis: Neural Networks and Model Comparison Due: Oct 27, in class Instructions: 1. Continue the analysis we started in PR1. A few reminders: Make sure you entered Prior Probability information (p. 39) and Profit Matrix information (p. 40) correctly. Make sure you impute the missing values as instructed on p. 104. Make sure you add the Log() transformations as instructed on pp. 116, 119, and 120. Then follow the analysis steps described in the Getting Started with SAS Enterprise Miner text, pp. 125-139 to add a variable selection node, fit a default neural network model, fit an AutoNeural model, and finally add a model comparison node to compare your four models (Decision Tree, Logistic Regression, Neural, AutoNeural). 2. Turn in a printed word document with (i) a brief one-paragraph description of the project problem, (ii) a brief one-paragraph outline of your analysis including a
Background image of page 1
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: screen shot of your final Diagram as shown on p.136, and (iii) screen shots of various analysis steps as listed below. Screen Shots: 1. A variable importance plot, as shown on p. 126 (top-left corner plot only) 2. An ROC chart that includes the four models, plus the baseline, as shown on p. 137 3. A cumulative Lift chart for the four models, as shown on p. 137 4. A profit chart for the four models, as shown on p. 139. Expand the plot to make all the details visible, as shown on p. 139. Include both Training and Validation charts. Questions: 1. Interpret the ROC charts shown on screen #2. Define the ROC Index. Which model appears to have the largest ROC index? How do you visually identify such a model by looking at the ROC charts? Hint: read p. 137. 2. Refer to screen #4. How many bins should we target so that marginal profit (=profit per customer) is maximized? Is this always a good marketing strategy? Why, or why not?...
View Full Document

{[ snackBarMessage ]}

Ask a homework question - tutors are online