This study describes the use of the K Nearest Neighbor method for data classification of patients with heart failure . In this research , several stages and methods were carried out . At the preprocessing stage of this study using the MinMax Scaler method for normalization and the LogX method to reduce the level of skewness and outliers . At stage s k enario test in research is using methods Cross Fold Validation with value k = 3, 5, 7 and 9 and as a differentiator in the research is carried out two processes are different , namely the implementation of the K-NN using techniques SMOTE earlier and implementation K- NN without SMOTE Technique . In measuring the classification results focus on the level of accuracy and precision . In a study of this data is used amounted to 299 patients who come from UCI Machine Learning . Measurement of the distance to the K-NN in the research is to use vector Euclidean Distance with a value of neighborhood 3, 5, 7 and 9. From the implementations are done obtained results that classification K Nearest Neighbor to the data of patients failing heart without SMOTE obtained results validate the best there is on the Nearest Neighbor k = 9 with an average accuracy of 71.59% and accuracy is the highest that is by 83.33% in the fold second . Whereas the highest accuracy test data was obtained at Nearest Neighbor k = 3 with an accuracy value of 71.66% . classification K Nearest Neighbor to the data of patients failing heart by using SMOTE obtained results validate the best there is on the Nearest Neighbor k = 3 with an average accuracy of 87.88% with accuracy the highest that is by 80.14% in the fold sixth and eighth . Whereas in the test data the highest accuracy was obtained at Nearest Neighbor k = 3 with an accuracy value of 63.33%.

Keyword: Classification , K-NN, Heart Failure , SMOT E , Cross Fold Validation.