MODERN REGRESSION METHODS FOR PREDICTIVE BUSINESS ANALYTICS
Date/Time : Monday, Aug 18, 8 am - Noon Fee : $300 ($350 after July 31)
In this half day course, we shall focus the challenges and issues associated with applying regression methods to big data problems in business. We shall focus on both:
Explanatory modeling: the process of building and applying a statistical model that is interpretable; and
Predictive modeling: the process of building and applying a statistical model to data in order to predict new or future observations.
We shall follow Arnold Zellner’s advice that the best explanatory models are “sophisticatedly simple”. Techniques to be discussed include multiple adaptive regression splines (MARS), shrinkage methods including lasso and elastic net and marginal model plots.
Understand the pros and cons of subset selection versus shrinkage methods approach (including lasso and elastic net) to model selection
Become familiar with the use of multiple adaptive regression splines (MARS) in analyzing regression data
Understand the use of marginal model plots to decide whether the regression model or the logistic regression model under consideration is a valid one or not
Become familiar with the situations in which quadratic and/or interaction terms need to be included in a logistic regression model
Professor Simon Sheather
Dept. of Statistics, Texas A&M University
Professor Sheather has over 20 years of experience applying analytics and statistical methods in business. His clients inclueed banks, biotechnology, hospitality service companies, fashion, transportation, real estate, consumer products and government. During this time he published over 75 papers and 2 books. He is listed on the ISIHighlyCited.com website among the top one-half of one percent of all mathematical scientists for citations of published work.
Date/Time : Monday, Aug 18, 1 - 5 pm Fee : $300 ($350 after July 31)
Ratings can tell you how a customer feels about your product or service, but text data provides the "why." A political poll can monitor a candidate's popularity, but only converstations with voters can explain the changes.
Customer ratings alone, cannot explain why customers are unsatisfied or satisfied with a product or service. Answering the why in customer relations is key to managing an organization and improving quality. Traditionally customer focus groups are used to examine customer issues. Used effectively, they can provide answers, but at high cost and since they involve small customer groups, they may not reflect the broader market.
Today there is another approach. One that is more accurate and provides the answer to why more quickly. An approach that lets organizations incorporate the opinions of thousands of customers, rather than just a few. Today, we can ask for customer opinions and analyze them using text analytics.
If customers have problems with your product or service, we can examine their comments to see whether this is a new event or something that started some time ago.
This workshop describes the principles used for computer analysis of written information, in particular customer comments, and how to integrate these with other data to build a complete quality assurance system. A system that not only answers what but why. Although examples are illustrated using SAS Text Miner, the general approach is software independent.
1 pm - 2 pm : Text Analytics - History and Terminology
2 pm - 3 pm : Measuring Customer Satisfaction using Text Data
3 pm - 4 pm : Text Analytics for Building QA System
4 pm - 5 pm :
A Case Study - Managing an Integrated Quality Assurance System.