Customer Churn Reduction Based on Action Rules and Collaboration

Introduction

Customer churn, also known as customer attrition, refers to the loss of existing customers who cease the relationship with an organization in a period of time (Renjith, 2017; Jain & Jana, 2021). Customer churn leads to lower volume of service consumption, reduced product purchase, less customer referrals. Furthermore, the cost of acquiring a new customer is much higher than the cost of retaining an existing customer (Siber, 1997). Reducing customer churn can be significantly beneficial (Van den Poel & Lariviere, 2004; Kowalczyk & Slisser, 1997). For example, in financial services, a 5% increase in customer retention produces more than a 25% increase in profit (Reichheld and Detrick, 2003).

It is imperative to build a recommender system that can provide effective recommendations to reduce customer churn. Recommender system is a subclass of information filtering systems that seek to predict the rating or preference a user would give to an item (Ricci, Rokach, & Shapira, 2011). According to the techniques applied, recommender systems are categorized to - collaborative filtering recommender systems, content-based recommender systems, demographic recommender systems, knowledge-based recommender systems, community-based recommender systems, and hybrid recommender systems (Su & Taghi, 2009). The process of building a knowledge-based recommender system is facilitated with the use of knowledge base, which contains data about rules and similarity functions to use during the retrieval process (Jannach, Zanker, Felfernig, & Friedrich, 2011). It relies more on the domain knowledge, by utilizing the expert knowledge to decide which item to recommend, and to what extent this item is meaningful and useful to the user.

Action rule mining is one of the technologies that have been successfully applied in building knowledge-based recommender systems to address the customer churn issue (Tarnowska & Ras, 2019; Tarnowska, Ras, & Daniel, 2020). In a knowledge-based recommender system based on classification rules and action rules, action rule mining is a major step in extracting knowledge in the process of recognizing the recommendations. The quality of action rules determines the effectiveness and coverage of recommender systems. Normally, support and confidence are used to measure the quality of discovered action rules (Ras & Wieczorkowska, 2000; Tzacheva, Sankar, Ramachandran, & Shankar, 2016). In practical applications, action rules are regarded as interesting only if their support and confidence exceed the predefined threshold values. Moreover, if an action rule has a large support and high confidence, it indicates that this action can be applied on a large portion of customers with a high chance (Ras & Tsay, 2003) to be successful. To increase the efficiency of knowledge-based recommender systems which use action rule mining for reducing customer churn, it is necessary to improve the quality of discovered action rules. However, there is little research work done which focuses on improving the quality of discovered action rules.