This paper discusses the performance of the K-Nearest Neighbor Algorithm with dimension reduction using Principal Component Analysis (PCA) in the case of diabetes disease classification. A large number of variables and data on the diabetes dataset requires a relatively long computation time, so dimensional reduction is needed to speed up the computational process. The dimension reduction method used in this study is PCA. After dimension reduction is done, it is continued with classification using the K-Nearest Neighbor Algorithm. The results on diabetes case studies show that dimension reduction using PCA produces 3 main components of the 8 variables in the original data, namely PC1, PC2, and PC3. Then classification result using K-Nearest Neighbor shows that by choosing 3 closest neighbor parameters (K), for K = 3, K = 5, and K = 7. The result for K = 3 has an accuracy of 67,53%, for K = 5 had an accuracy is 72,72%, and for K=7 had an accuracy of 77,92%. Thus, it was concluded that the best accuracy performance for the classification of diabetes was achieved at K=7 with an accuracy of 77.92%.