Predicting Potential Customers in Direct Marketing Using Uplift Modelling and Advanced Machine Learning

Uplift Modeling can be an effective machine learning method when identifying the potential customers who have the highest possibility of creating a positive impact on the marketing strategies. Unlike traditional approaches that are solely based on predicting outcomes, the Uplift model aims to understand the causal relationship between marketing strategies and customer responses. In this paper, we examine how machine learning can be effectively used in the product marketing stage. Using the Criteo uplift dataset, the study tries to solve the problem using the Uplift modelling concept to predict potential customers who react positively to direct marketing campaigns the study used three different approaches which combine with parallel computing: 1) Dual Uplift modelling with hypercomputing integration, 2) eXtreme Gradient Boosting (XGBoost), and 3) combining both the Dual Uplift model with XGBoost model. Once the model is developed, using the model, the study extends its scope to develop a web application using R shiny apps. These models were evaluated using statistical measures and discussed the business implications.