No abstract

A c c e p t e d M a n u s c r i p t  A novel profit-based feature selection method for churn prediction with SVM is presented.  A backward elimination algorithm is performed to maximize the profit of a retention campaign. Our experiments on churn prediction datasets underline the potential of the proposed approaches. *Highlights (for review)Page 2 of 36 A c c e p t e d M a n u s c r i p Effective churners Outflow New customers InflowConsists of N customers with average customer lifetime value CLV.The cost of contacting a customer is f.The cost of an incentive offer is dWould-be churnersClassified as churners (η η η η)Would-be churnersClassified as nonchurners (1-η η η η)Customer Churn Management Campaign AbstractChurn prediction is an important application of classification models that identify those customers most likely to attrite based on their respective characteristics described by e.g. socio-demographic and behavioral variables.Since nowadays more and more of such features are captured and stored in the respective computational systems, an appropriate handling of the resulting information overload becomes a highly relevant issue when it comes to build customer retention systems based on churn prediction models. As a consequence, feature selection is an important step of the classifier construction process. Most feature selection techniques; however, are based on statistically inspired validation criteria, which not necessarily lead to models that optimize goals specified by the respective organization. In this paper we propose a profit-driven approach for classifier construction and simultaneous variable selection based on Support Vector Machines. Experimental results show that our models outperform conventional techniques for feature selection achieving superior performance with respect to business-related goals.

We study the assortment optimization problem under the Sequential Multinomial Logit (SML), a discrete choice model that generalizes the multinomial logit (MNL). Under the SML model, products are partitioned into two levels, to capture differences in attractiveness, brand awareness and, or visibility of the products in the market. When a consumer is presented with an assortment of products, she first considers products in the first level and, if none of them is purchased, products in the second level are considered. This model is a special case of the Perception-Adjusted Luce Model (PALM) recently proposed by Echenique, Saito, and Tserenjigmid (2018). It can explain many behavioral phenomena such as the attraction, compromise, similarity effects and choice overload which cannot be explained by the MNL model or any discrete choice model based on random utility. In particular, the SML model allows violations to regularity which states that the probability of choosing a product cannot increase if the offer set is enlarged.This paper shows that the seminal concept of revenue-ordered assortment sets, which contain an optimal assortment under the MNL model, can be generalized to the SML model. More precisely, the paper proves that all optimal assortments under the SML are revenue-ordered by level, a natural generalization of revenue-ordered assortments that contains, at most, a quadratic number of assortments. As a corollary, assortment optimization under the SML is polynomialtime solvable. This result is particularly interesting given that the SML model does not satisfy the regularity condition and, therefore, it can explain choice behaviours that cannot be explained by any choice model based on random utility.