Article Preview
Top1. Introduction
MNP or otherwise known as “customer churn in disguise”, is the complete churning of a subscriber from one service provider (SP) to a supposedly better SP without changing the phone number (Lin, Chlamtac, & Yu, 2003). If this was not available, customers of any service provider would have to forgo their existing mobile number (MN) and take a new MN and it could come with the cost of a new business card or loss of valuable number if no backup exists. This is an excellent opportunity for many well-known business customers who may be reluctant to churn their SP and give up their mobile number even if the alternative SP provides better services (Gans, King, & Woodbridge, 2001). MNP reduces switching costs and increases telecom competition, thereby promoting sustainability by encouraging operators to offer innovative solutions (Nwankwo & Njoku, 2020).
However, telecom companies operating in saturated markets are continuously faced with the challenges of retaining and sustaining their potential customers due to the MNP techniques, which is not a blessing in disguise. This is because these companies expend enormous resources to attract and retain customers. They understand that customer retainership is the panacea for preventing customers from churning. Moreover, not only is anticipating churn difficult, but trying to predict how significant a problem it will turn out to be; how long it will last; and the ultimate consequences (in terms of loss of revenue, especially when a high potential customer churns to a better competitor) it will incur on the company is a challenge (Verbeke, Dejaeger, Martens, Hur, & Baesens, 2012). It is also noteworthy that mobile number churning can result in difficulties distinguishing different network operators’ customers’ number patterns when placing a call. Notably, doing the same was without difficulties prior to the introduction of churning.
Several machine learning techniques have been proposed in recent years to aid the discovery of hidden patterns in datasets, modelled to resolve the problems posed by churn across the telecom operators that have led to customers’ decision to change their mobile phone numbers and their service providers respectively. Some of these machine learning techniques include hybrid firefly (Amin et al., 2017), ProfTree (Höppner, Stripling, Baesens, vanden Broucke, & Verdonck, 2018), rough set theory, RotBoost (De Bock, & Van den Poel, 2011), logit leaf model (De Caigny, Coussement, & De Bock, 2018), multilayer perceptron neural network (MLP), Decision Tree C4.5, support vector machines (SVM) (Huang et al., 2010; Huang, Kechadi, & Buckley, 2012), random forest and particle swarm optimization (Idris, Rizwan, & Khan, 2012), Bayesian network (Kisioglu, & Topcu, 2011), and SVM-Polynomial kernel (SVM-POLY) (Vafeiadis, Diamantaras, Sarigiannidis, & Chatzisavvas, 2015). However, little or no research work has been published on solving the real problem of mobile number portability, which could be regarded as another dimension of customer churn in disguise. Recently, there has been an increase in number portability, and in Nigeria, for example, several advertisements were promoted by mobile telecom operators on the issue of portability in late 2018 and 2019. One of the most cited benefits is the flexibility it offers to potential or prospective customers, particularly in terms of retaining their telephone number following porting to a new platform.