Improving Customer Churn Prediction: A Study of Counterfactual Explanations Using Wachter's Method, Growing Spheres Method, and Genetic Algorithms

Customer churn prediction is critical for businesses to retain customers and maintain revenue. In this paper, we compare three counterfactual explanation methods: Watcher's Method, Sphere Method, and Genetic Algorithm, which are applied to customer churn prediction models. We evaluate the effectiveness of these methods in generating counterfactual instances, aiming to understand the minimal changes required to alter a customer's churn prediction. The key metrics, including Fischer Score, Imbalance Ratio, Volume, and Number of Features, are analysed to assess the performance and interpretability of the models. Our findings highlight the strengths and weaknesses of each method, providing insights into their practical applications for improving customer retention strategies.