Frequent uses of information technologies will accumulate a large number of documents. If these documents are not classified, it will be difficult for the users to find information in a large unorganized set of documents. Large human resources will be needed if the documents are classified manually. How to analyze and classify such documents efficiently and automatically is the goal of this research. In this thesis, we used Hadoop platform to classify documents. At First, we analyze the documents to find representative keywords for each document. We then compute the TF-IDF weights of each keyword. After that, we use Hierarchical Agglomerative Clustering and K-Nearest Neighbors Algorithm to classify documents. After classification, we use Concept Analysis to describe the relationship between keyword concepts and to build an ontology. Finally, we use the ontology to improve the accuracy of the classification.