Utilizing machine learning to project the nancial outcomes of reconnecting with potential customers of the same industry

Abstract

In competitive markets, it is costly to attract new customers in the business as they already have a wide customer base; that being the case, businesses spend a healthy budget to bring back customers who have once been with them. Despite the business investing heavily in trying to retain their customers, the re-engagement of the customers is not satisfactory as many businesses use intuition, experience, and traditional methods for marketing literature. Moreover, there is a global pandemic (COVID-19) that is hampering businesses everywhere. While the majority of the businesses are operating on a loss, a few small businesses are already being shut down. Thus, there is a need to increase the pro tability of the businesses and retain their customer base. To increase customer turnover and re-engagement, this paper focuses on the implementation of intelligent business practices using machine learning and neural networks. The paper focuses on analyzing the customer behavior based on their purchase behavior and transactional values for customer retention. The classi cation is done to identify the customers who are pro table to the business and customers who are likely to churn due to various reasons. In this paper, we proposed a Multi-layer perceptron (MLP) to segment the customers according to RFM methodology and customer pro tability index. The result of MLP was compared with K-Nearest Neighbors (KNN) and Support Vector Machine (SVM) as the later models have been widely used. Additionally, Bidirectional Long Short-Term Memory (LSTM) models have been implemented for primary customer classi cation and sales prediction. The prediction model is an attempt to reduce - nancial loss on marketing campaigns for re-engagement of customers in the business.