The retail industry plays a crucial role in the global economy, driving innovation and consumer satisfaction through efficient operations and effective inventory management. The aim of this research is to develop and optimize a combination between item-based and user-based collaborative filtering recommendation system for smart vending machines. Specifically, due to the inability to obtain consumer ratings for products during the sales process on retail machines, this study utilizes historical sales data from smart vending machines. The daily average sales of different products are used as implicit rating indices to more accurately reflect consumer preferences in the modeling process. The study employs k-fold cross-validation and evaluates the predicted ratings using root mean squared error (RMSE), mean squared error (MSE), and mean absolute error (MAE) metrics. Additionally, the recommendations or recommended items were assessed using important metrics to evaluate the confusion matrix. The results indicate that the proposed recommendation system demonstrates reliability and effectiveness. This study contributes to the technological advancement of the retail industry by developing a robust combined model tailored for smart vending machines. The proposed model can significantly improve customer satisfaction and operational efficiency in vending machine management.