Machine learning (ML) refers to the study of methods or algorithms designed to learn the underlying patterns in the data and make predictions based on these patterns. A key characteristic of ML techniques is their ability to produce accurate out-of-sample predictions. We review two popular machine-learning methods – decision trees and Support Vector Machines (SVM) in detail.
Browse by title
Daria Dzyabura and Hema Yoganarasimhan
Vildan Altuglu and Rainer Schwabe
Litigation presents significant challenges involving the identification, sorting, and analysis of large amounts of data. Machine learning, which utilizes algorithms and systems that improve their performance with experience to classify information and to make predictions, is well-suited to these tasks. In this chapter, we discuss current machine learning applications in legal practice, as well as some potential applications of these techniques in support of expert witness testimony in commercial litigation.
Murali K. Mantrala and Vamsi K. Kanuri
We survey the methods, advances, and insights from research and applications pertaining to Marketing Optimization Methods over the past 70 years. Specifically, we classify extant marketing optimization problems into two key typologies based on: (1) the number (“single” or “multiple”) of “sales entities” and marketing input variables involved in the problem, and (2) the nature of the objective function (e.g., static or dynamic). We discuss the modeling and solving of optimization problems that fall under these typologies. In each example, we summarize the problem; the choice variables; the constraints; the sales response model; the objective function; the solution approach/technique; and optimization insights/principles from the solution.
Keiko I. Powers
This chapter uses multivariate time-series methods to study one of the most serious public policy problems, the fight against narcotics abuse. The effects of methadone treatment and legal supervision of narcotics use and criminal activities were assessed by applying cointegration and error correction methods that disentangle the long-term (permanent) and the short-term (temporary) effects of intervention. Overall, the system dynamics among these variables were characterized by long-term rather than short-term relationships. Methadone maintenance treatment demonstrated long-term benefits by reducing narcotics use and criminal activities. Legal supervision, on the other hand, did not reduce either narcotics use or property crime in the long run. The chapter explores the policy implications of these findings.
Donald R. Lehmann
This chapter discusses important methods and issues in using meta-analysis to develop a knowledge base in marketing. After defining meta-analysis and explaining its role in marketing, the author discusses various steps in a meta-analytic study, focusing both on design and statistical issues. He then presents a comprehensive tabular overview of published marketing meta-analyses in various subfields of marketing.
John Roberts and Denzil G. Fiebig
This chapter examines the use of choice models in marketing. After briefly describing the genesis of choice modeling, we introduce the two basic workhorses in choice modeling, the logit and probit models. We use these two models as a platform from which to show how additional phenomena can be introduced, including multistage decision processes, dynamic models, and heterogeneity. After a description of some more advanced models, we close by illustrating how these models may be used to provide insight to marketing managers by discussing a number of choice modeling applications.
In this chapter I present three techniques—Cluster analysis, factor analysis, and multidimensional scaling—popular with marketing researchers and consultants because they help achieve frequently encountered marketing goals. Cluster analysis is useful in finding customer segments, factor analysis is useful for survey research, and multidimensional scaling is useful in creating perceptual maps.
Jorge Silva-Risso, Deirdre Borrego and Irina Ionova
We develop a consumer response model to evaluate and plan pricing and promotions in durable good markets. We discuss its implementation in the US automotive industry, which “spends” about $50 billion each year in price promotions. The approach is based on a random effects multinomial nested logit model of product and transaction-type choice. Consumers differ in their overall price sensitivity as well as in their relative sensitivity to alternative pricing instruments which has to be taken into account to design effective pricing programs. We estimate the model using Hierarchical Bayes methods to capture response heterogeneity at the local market level. We illustrate the model through an empirical application to a sample of data drawn from J.D. Power transaction records.
Zoë Chance, Ravi Dhar, Michelle Hatzis, Michiel Bakker, Kim Huskey and Lydia Ash
In this chapter, we share the 4Ps Framework for Behavior Change, designed to organize research findings to make them more easily applicable in the real world. We offer levers the well-meaning planner can employ to support the healthy intentions of others, and share examples of how the 4Ps Framework is being applied at Google. Although our examples focus on nudging people toward healthy food choices, similar strategies can be used to nudge people’s behavior in any direction that supports their own intentions. We offer advice for influence one-time decisions via (1) the combination of choices offered, (2) the choice environment, and (3) communication about the choices. We also offer advice on supporting individuals in the development of good habits, to make better choices in any time or place.