Post-Conference Data Mining Workshop
Wednesday, March 25
1:30 - 5:00 p.m.
This post-conference, half day workshop on data mining offers another opportunity to expand your SAS data mining and predictive modeling skills. Stay later to take advantage of exclusive training opportunities. This workshop is designed for marketers, data analysts, business analysts and executives who want to understand the best practices of conducting end-to-end analytics – from data preparation to exploratory analysis to predictive modeling and implementation. This workshop introduces and demonstrates the introductory and advanced features available in the SAS® Enterprise MinerTM and SAS® Text MinerTM. Experience with either product is not required to attend. The $150 fee includes a boxed lunch and afternoon snacks.
-
SAS Enterprise Miner 6.1 in Action
Wayne Thompson, SAS Institute
Enterprise Miner is a modern, distributed client-server system for predictive model development. The system enables wide area collaboration on data mining projects and extensive integration and configuration opportunities. This demonstration session will focus on developing classification models using the Sample, Explore, Modify, Model, Assess process. The tasks of exploratory statistical and visualization discovery, variable transformations and selection, filtering outliers, model validation and scoring will be covered using the new EM 6.1 SAS 9.2 release.
-
Predicting Segments from a Market Research Study Using SAS Enterprise MinerTM
Randall S. Collica, Hewlett Packard Co
One of the holy grails of marketing research is to understand the attitude your customers have towards your company and its products and services that are offered. To accomplish this end, a typical method is to perform a market research survey from a representative sample of selected customers and ask them a set of questions that will directly facilitate a common set of attitude groups or segments given certain combinations of responses to the questions. Once completed, the researcher now has each response coded into a particular attitudinal segment and each segment has a particular profile of customer demographics. This data is useful for understanding how your customers feel, their attitudes, and can give valuable insight for developing specific targeted marketing or sales campaign tactics. However, at this point this analysis is confined only to the survey respondents.
This presentation describes how one can create a model that will accurately score all customers on the database from which the sa mple was derived and thereby have the market research extended to all customer rather than just the survey recipients? The paper will present one such tactic using SAS Enterprise MinerTM to develop and build a scoring model. Analysis of prediction confidence intervals is also shown using a bootstrap algorithm to estimate the confidence intervals of predicted probabilities.
-
A Demonstration of Text Mining Using the Voluntary Reporting of Adverse Events
Pat Cerrito, Professor, University of Louisville
We demonstrate how text mining can be used to examine actual medication risk instead of relying upon perceived risk or historical risk as is currently done when investigating adverse events. We investigate the voluntary reporting of adverse events resulting from medications and vaccinations using SAS Enterprise Miner and SAS Text Miner. These reporting databases are publicly available and can be downloaded from the web in ascii format. The VAERS (vaccine adverse events) report has a text column that allows for a free-form description of the adverse event. The AERS (medication adverse events) requires standardized terms, with each term an observation and multiple observations (and terms) per event. Both databases include patient demographics and outcomes. SAS Text Miner can use both types of text fields to investigate the adverse event and to examine patterns of occurrence of side effects, which can then be compared to patient outcomes.
Note: You may register for this workshop when you register for SAS Global Forum.