The paper presents a movie genre recommendation system tailored for small and medium-sized enterprises (SMEs) using imbalanced data and adaptive classification costs. It leverages machine learning algorithms, particularly ensemble methods like Gentle Boost and AdaBoost, to effectively predict genre preferences based on a detailed user profile dataset without relying on movie-specific information. The proposed system addresses challenges such as imbalanced data distribution and varying classification costs, providing a more accurate predictive model for targeted marketing in the movie rental industry.