The paper discusses applying RFM-D analysis for customer segmentation using modern machine-learning methods. The authors focus on integrating three main approaches: reinforcement learning methods, unsupervised ML methods, and supervised ML methods. The first approach, based on reinforcement learning, allows to adaptively adjusting segmentation strategies based on variable customer characteristics and preferences and can continuously improve the process of selecting optimal strategy to achieve the objective function, which is essential for businesses operating in a dynamic market environment. For unsupervised learning, the K-means clustering method is considered, which helps to identify more homogeneous groups of customers based on characteristics such as purchase history, frequency and other factors, allowing businesses to customize their unique customer offers and related marketing strategies more accurately. At the same time, machine learning methods, such as classification algorithms, can predict customer behavior based on training data, ensuring high prediction accuracy and improving the quality of management decisions regarding unique offers to each customer segment. The study results show that combining these approaches allows for effective customer segmentation and optimized customer interaction, critical for increasing loyalty and overall efficiency of business strategies.