Over all 7,000 languages around the world, nearly one third of them are at the risk of extinction while some others may suffer from other vulnerable dangers. Global efforts are needed to conserve the human language and culture. In this paper we selected the features with suggested causes of the danger of those languages from the Cambridge Handbook of Endangered Languages, and the actual data were obtained from various authoritative sources. We analyzed the data to check which features contributing the classification of endangered languages, we found that GDP, latitude and longitude were associated with language endangerment. Finally, we applied decision tree algorithm and its ensemble learning algorithms including Bagging, RandomForest and AdaBoosting for classification. We compared the performance of the four algorithms, and our evaluation indicates that ensemble algorithms overall have better higher accuracy rate compared to the base classifier, decision tree.