This study aims to develop a machine learning-based construction material stock prediction model using a hybrid approach that combines K-Means Clustering as a sales pattern grouping method and Support Vector Machine (SVM) as a classification method to predict material sales levels. This research was motivated by the problem of stock management at Toko Usaha Jaya in Lubuklinggau City, which is still done manually, thus potentially causing excess stock that increases storage costs and stock shortages that can lead to lost sales opportunities and decreased customer satisfaction. The data used includes material names, initial stock quantities, quantities sold, remaining stock, and selling prices collected during the period from January to December 2023. The results show that the hybrid model is capable of grouping materials into three categories, namely very popular, fairly popular, and less popular, with a Silhouette Score of 0.42, indicating fairly good clustering quality. Furthermore, the SVM model produced a classification accuracy rate of 99%, reflecting an increase in stock prediction accuracy compared to manual management methods. These findings indicate that the application of the K-Means and SVM hybrid model can improve inventory management efficiency and support more accurate and effective data-driven decision making.