Classification is a process that automatically places text documents into a text based on the content of the text. Classification can help us classify many text documents that have been published, with the classification, these text documents can be reached easily and quickly. Feature selection can be used to improve the performance of text classification in terms of learning speed and effectiveness. In the Chi-Square feature selection experiment, a 1% threshold combination with a parameter value of k=6 is the combination chosen to be the best model. In testing the new data, the K-Nearest Neighbor model by selecting the Chi-Square feature produces precision performance, recall, F1-Score, and accuracy respectively, namely 85%, 83.3%, 88.2%, and 92.3%. In the Gini Index feature selection experiment,1% threshold combination with a parameter value of k=4 is the combination chosen to be the best model. This threshold selects about 31 features with the highest Gini Index value. In testing the new data, the K-Nearest Neighbor model by selecting the Gini Index feature produces precision performance, recall, F1-Score, and accuracy respectively, namely 81.2%, 80.3%, 81.6%, and 86.6%.