This study aims to classify reject patterns in the production process using the K-Means Clustering method. The dataset consists of 870 records collected from the production line, containing information such as product name, reject type, process stage, and production quantity. Through a data mining approach, data preprocessing steps such as cleaning, encoding, and normalization were performed prior to the clustering process. The Elbow Method indicated that the optimal number of clusters is three. Each cluster exhibits distinct characteristics: light rejects with small quantities in early stages, heavy rejects with large quantities, and moderate rejects with random distribution. These findings are expected to assist management in formulating more targeted strategies for process improvement and quality control. By identifying common reject patterns within each cluster, companies can adopt a more proactive approach to minimizing production defects and enhancing overall operational efficiency.