Uplift modeling is a type of predictive modeling that estimates the incremental effect of performing some action on customer behavior. This allows for a high degree of customization and improved targeting selection in marketing campaigns with the aim to maximize the effect of the campaign and the returns on marketing investment. In previous work, we have extended the maximum profit measure towards uplift modeling for identifying the use in terms of profitability of an uplift model.
In this paper, we take it one step further and consider the profit maximization objective during the model construction phase. We focus on adapting existing uplift decision trees, which in turn are direct generalizations of standard decision trees such as CART and C4.5. Inspired by the literature on cost-sensitive learning techniques, we develop and evaluate an array of approaches to replace the impurity measure in the splitting criterion of decision trees by including components of the maximum profit measure. We present the results of an experimental study, comparing the performance of the various proposed implementations with existing approaches such as uplift random forests on two real-world datasets. To evaluate, we adopt both profit and qini-based model performance assessment approaches.
Original languageEnglish
Title of host publicationEURO 2018: 29th European Conference on Operational Research
PublisherEuro: The Association of European Operational Research Societies
Pages193-193
Number of pages1
Publication statusPublished - Jul 2018
EventEURO 2018: 29th European Conference on Operational Research - Valencia, Spain
Duration: 8 Jul 201811 Jul 2018
Conference number: 29
http://euro2018valencia.com/

Conference

ConferenceEURO 2018
Abbreviated titleEURO 2018
CountrySpain
CityValencia
Period8/07/1811/07/18
Internet address

ID: 38822300