Courses
  /  
Descriptions
IEMS 462-1: Predictive Analytics I

Quarter Offered

Winter : TTH 2-3:20 PM ; Tamhane

Prerequisites

A course in statistics covering through linear regression and a course in matrix algebra

Description

This course is Part I of a two-part sequence in predictive analytics. Part I will cover classical parametric regression and classification models while Part II will cover modern nonparametric models. The course will begin with a brief review of simple linear regression and correlation and then proceed to a detailed discussion of multiple regression. Next it will cover binomial and multinomial logistic regression and classification. Discriminant analysis will be offered as an alternative technique to logistic regression. The final topic will be generalized linear models including Poisson regression for count data and Cox proportional hazards model for survival data. The underlying theory of these methodologies will be covered in detail and will be illustrated by fitting models to medium and large data sets.

LEARNING OBJECTIVES

  • Students will the learn the theory and predictive analytics applications of parametric regression based methods including multiple regression, logistic regression, generalized linear models and Cox proportional hazards model.
  • Students will develop capability in writing R codes to fit above models.

TOPICS

  • Simple linear regression and correlation (Review)
  • Multiple linear regression: Basic
  • Multiple linear regression: Model diagnostics
  • Multiple linear regression: Variable selection and model building
  • Logistic regression and classification
  • Discriminant analysis
  • Generalized Linear Models
  • Survival Analysis

MATERIALS

Draft of the book "Parametric Regression and Classification Models for Predictive Analytics" by A. C. Tamhane and E. C. Malthouse.

Syllabus