Towards explainable direct marketing in the telecom industry through Hybrid machine learning

Abstract

Direct marketing enables businesses to identify customers that could be interested in product offerings based on historical customer transactions data. Several machine learning (ML) tools are currently being used for direct marketing. However, the disadvantage of ML algorithmic models is that even though results could be accurate, they lack relevant explanations. The lack of detailed explanations that justify recommendations has led to reduced trust in ML-based recommendations for decision making in some critical real-world domains. The telecommunication domain has continued to witness a decline of revenue in core areas such as voice and text messaging services which make direct marketing useful to increase profit. This paper presents the conceptual design of a machine learning process framework that will enable telecom subscribers that should be targeted for direct marketing of new products to be identified, and also provide explanations for the recommendations. To do this, a hybrid framework that employs supervised learning, case-based reasoning and rule-based reasoning is proposed. The operational workflow of the framework is demonstrated with an example, while the plan of implementation and evaluation are also discussed.