The aim of the present paper is to suggest a neutrosophic set model for information retrieval and to develop methods and algorithms for information retrieval based on the neutrosophic set model. A neutrosophic set has the possibility of being a common structure for the vagueness analysis of data sets also including big data sets. Here we defined useful techniques like distance and similarity between two neutrosophic sets that have been applied to document classification in information retrieval. We define an innovative algorithm for classifying documents based on Euclidian distance between two neutrosophic sets. About 2500 documents in seven different categories are used for evaluation this new algorithm. Our experiments show that Neutrosophic Classification Algorithm achieved 95% performance.