Uplift Modeling in E-Grocery

Uplift modelling contains a set of techniques that allow predicting a change in an individual's behaviour as a result of some form of targeted influence, such as marketing communication. In the digital marketing industry, there are many ways to target customers, and Uplift modelling is used to select the customers through whom target marketing will be most effective. Uplift modelling increases customer retention and helps companies optimise their marketing budget. Uplift modelling is a task at the edge of Causal Inference and Machine Learning. This paper presents a practical application of Uplift modelling in the fast-growing field of E-grocery. A detailed comparative analysis of the main Uplift modelling approaches: SoloModel, TwoModel, the Class Transformation and uplift modelling directly is made in the paper. As a result of experiments carried out on three different data sources, the optimal approach for solving the Uplift modelling problem has been identified. Also, this paper presents the design of the experiment to perform Uplift modelling. Translated with www.DeepL.com/Translator (free version)