International Tourism is one of the most important source of economical revenues for some countries. For that reason, building a sentiment analysis model using recurrent neural network (RNN) to analyze the international tourist’s opinion on twitter was carried out. The dataset that is used in this study is using 1653 data that consists of tweets.Then, the obtained data was preprocessed. The preprocessing step consists of data cleaning, casefolding, stop word removal, and tokenize the data. The data was labelled by using texblob. Which resulting 1523 data labelled positive and 130 data labelled negative. Because of the imbalance number of positive data and negative data, oversampling was applied in order to decrease sparse between positive data and negative data. The classification process is carried out using RNN method by using keras, a python library. The evaluation is carried out by comparing different training and testing data ratios, also different batch sizes for each training and testing data ratio. The highest accuracy that was obtained in this research is 79.86% by using 75% of the data for training data and 25% of the data for testing with the batch size set to 16.