Teaching Data Mining and Predictive Analytics to Undergraduates

• Activity Number: 414
• Type: Topic Contributed
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Education
• Author(s): Brant Deppa*+
• Companies: Winona State University
• Keywords: Data mining, prediction, classification, clustering, R

Abstract

The second applied statistics course often covers topics such as multiple linear and logistic regression. Traditionally, courses in analysis of variance, experimental design, multivariate analysis, or nonparametric methods follow a regression course. This traditional progression may not be best for all students. For example, with the rise of big-data and the increased demand for "data scientists" in industry, a course covering data mining and predictive analytics may be better. While some of the computational and theoretical developments of these methods might be beyond the ability undergraduates, a good foundation in regression makes the discussion of modern predictive and classification methods such as classification & regression trees, neural networks, random forests, Bayesian classifiers, and support vector machines possible. As all of these methods are available in the open source statistical environment R, as well as other commercial software packages, students are able to apply, critique, and compare these methods for solving prediction and classification problems. The author will share his experiences in teaching such a course to undergraduates.