Customer segmentation plays a vital role in designing effective marketing campaigns to drive business growth and maximize revenue. In this study, we apply Principal Component Analysis and Autoencoder to extract meaningful features, reducing data complexity while preserving essential information. To enhance clustering accuracy, we first utilize DBSCAN to detect and remove noise before employing the K-Means algorithm for customer segmentation. This approach helps identify distinct customer groups, providing valuable insights into consumer behavior. Furthermore, we analyze potential customers who have the possibility of being upgraded from loyal customers to VIP status. Finally, we implement a Random Forest Classification model to predict potential new customers based on fundamental customer information, enabling businesses to develop proactive strategies for customer acquisition and retention.