Analysis of time series data requires some assumptions that stationary and homogeneity of variance. In many cases is rarely found time series data that satisfy those assumptions. Those are due to the complex non-linear relationship between the multidimensional features of the time series data. KNN method is one of the Learning Machine algorithm (LM) which is considered as a simple method to be applied in the analysis of data with many dimensions variable. This method can be used when it does not meet the classical assumptions. This study aims to see the performance of KNN and ensemble KNN. Although this method is simple but this method has advantages over other method. For instance, it can generalize from a relatively small training set. In This method is very important to choose the number of k-nearest neighbors. Ensemble technique is a method that has accuracy of prediction and efficiently used in the KNN method, so it is not necessary to search the optimal number k. The result shows that MAPE, MAE, and RMSE of prediction will be small if the number of k-nearest neighbors large. Overall, KNN ensemble method has a better performance than KNN method. 7994Dewi Sinta et al.